PreprintPDF Available
Preprints and early-stage research may not have been peer reviewed yet.

Abstract and Figures

Epidemics can negatively affect economic development except mitigated by global governance institutions. We examine the effects of sudden exposure to epidemic disease on human capital outcomes using evidence from the African meningitis belt. Meningitis shocks reduce child health outcomes, especially in periods when the World Health Organization (WHO) does not declare an epidemic year. These effects are reversed when the WHO declares an epidemic. Children born in meningitis shock areas during a year when an epidemic is announced are 10 percentage points (pp) less stunted and 8.2 pp less underweight than their peers born in non-epidemic years. We find suggestive evidence for the crowd-out of routine vaccination during epidemic years. We analyze data from World Bank projects and find evidence that an influx of health aid in response to WHO declarations may partly explain these reversals.
Content may be subject to copyright.
The Epidemic Effect: Epidemics, Institutions and
Human Capital Development
Belinda Archibong
Barnard College
Francis Annan
Georgia State University
Uche Ekhator-Mobayode
World Bank
June 5, 2022
Abstract
Epidemics can negatively affect economic development except mitigated by global
governance institutions. We examine the effects of sudden exposure to epidemic disease
on human capital outcomes using evidence from the African meningitis belt. Meningitis
shocks reduce child health outcomes, especially in periods when the World Health
Organization (WHO) does not declare an epidemic year. These effects are reversed
when the WHO declares an epidemic. Children born in meningitis shock areas during
a year when an epidemic is announced are 10 percentage points (pp) less stunted
and 8.2 pp less underweight than their peers born in non-epidemic years. We find
suggestive evidence for the crowd-out of routine vaccination during epidemic years.
We analyze data from World Bank projects and find evidence that an influx of health
aid in response to WHO declarations may partly explain these reversals.
JEL classification: I12, I15, I18, H84, O12, O19
Keywords: Epidemic, Disease, Vaccination, Aid, WHO, World Bank, Africa
Thanks to Ebonya Washington, James Fenske, S.K. Ritadhi, Marcella Alsan, Nancy Qian, Jerome Adda,
Kwabena Gyimah-Brempong, Ted Miguel, Erzo Luttmer and participants at the SEA, AEA, WGAPE,
NAREA and the pre-NBER conferences, and Columbia, USF and Delaware seminars for useful comments
and suggestions. Thanks to Sophie Danzig, Shristi Bashista, and Jessica Moreira for excellent research
assistance. The research benefited from conversations with World Bank employees, and we thank them for
helpful comments. We are grateful to Carlos Perez, Madeleine Thomson, Nita Bharti, and the WHO for the
data on meningitis used in this study. Errors are our own. Declarations of interest: none. This research did
not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Corresponding author. Barnard College, Columbia University. 3009 Broadway, New York, NY 10027,
USA. ba2207@columbia.edu.
1
1 Introduction
The virulence and human cost of recent epidemics have reignited policy debates around
optimal strategies to mitigate the economic burden of infectious disease. One of the most
rigorously debated policies is providing aid funding or stimulus to areas affected by epidemics.
How effective are these aid funding strategies in mitigating the negative effects of epidemics?
When countries are declared nationally epidemic by global health governance organizations
based on some threshold of infectious disease cases, this declaration may trigger an influx of
disaster aid and financing efforts that can improve human capital outcomes and reverse the
negative effects of epidemics. Our work provides key insights into this epidemic effect.
In this study, we ask two questions: (i) how do epidemics of infectious disease affect
human capital development? And (ii) what roles, if any, do global governance institutions
play in mitigating these impacts? Exploiting exposure to meningitis shocks and epidemic
years in the African meningitis belt, we assemble data on meningitis cases, epidemics, the
flow of the World Bank’s aid spending, and child health outcomes to investigate the effects
of epidemics on human capital outcomes. The meningitis belt comprises approximately 23
countries in Africa, extending from Senegal to Ethiopia and comprising over 700 million
individuals, that are frequently exposed to meningitis epidemics as shown in Figure 1a.
The epidemic1form of meningitis is caused by the bacterium Neisseria meningitidis and is
characterized by an infection of the meninges or the thin lining covering the brain and spinal
cord. Direct transmission is through contact with respiratory droplets or throat secretions
from infected individuals (LaForce et al., 2009; Garc´ıa-Pando et al., 2014). Infection is
associated with fevers, pain, reduced cognitive function, and- in the worst cases- permanent
disability and long-term neurological damage and death. Young children and adolescents
are particularly at risk of infection, and epidemics can be very costly for households, with
1Where epidemics are defined in the sub-Saharan Africa context as greater than 100 cases per 100,000
population nationally within a year by the WHO (LaForce et al., 2009).
2
households in the belt spending up to 34% of per capita GDP on direct and indirect costs
stemming from meningitis epidemics (Colombini et al., 2009).
We exploit quasi-random variation in district-level exposure to meningitis shocks and
country-year variation in the announcement of an epidemic year to examine these effects by
employing a panel regression framework. Our meningitis shock variable is constructed from
a new dataset of mean weekly meningitis cases per 100,000 population for districts across
eight countries in the belt from 1986 to 2008. The shock variable is an indicator that equals
one if meningitis cases in a given year are above the district’s standardized long-term mean,
following the definition of epidemics outlined by the World Health Organization (WHO,
2020)2.
We examine the effects of meningitis shocks on child health outcomes like stunting
or whether or not a child is underweight. The results on child health are economically
important, given the vast literature linking child stunting and underweight status, both
primary markers of malnutrition, with poorer cognitive and earnings outcomes in adulthood
(Jayachandran and Pande, 2017; Bisset et al., 2013)3. In other words, people who are shorter
and more underweight as children have worse health outcomes, lower cognitive ability and
lower earnings as adults. The results show that meningitis shocks or high, unexpected levels
of meningitis are associated with significant reductions in child health outcomes, reflected
in increased incidence of stunting and underweight status, particularly during non-epidemic
2The WHO defines an epidemic as “the occurrence in a community or region of cases of an illness clearly
in excess of normal expectancy. The number of cases indicating the presence of an epidemic varies according
to the agent, size, and type of population exposed, previous experience or lack of exposure to the disease,
and time and place of occurrence.” (WHO, 2020).
3A child is considered stunted or underweight if she has a height-for-age or weight-for-age (measured by
the height-for-age and weight-for-age z-scores) that is two standard deviations or more below the worldwide
reference population median for her gender and age in months. Jayachandran and Pande (2017) provides
a review of the literature showing a positive relationship between child height and adult height, with taller
adults having ‘greater cognitive skills, fewer functional impairments and higher earnings’. A recent literature
has also linked underweight status in early childhood to poorer cognitive outcomes later in life (Bisset et al.,
2013).
3
years. The effect of meningitis shocks on child health is nonlinear; meningitis shocks increase
child health outcomes during years declared by the WHO as epidemic years and reduce
health outcomes during non-epidemic years. Children born in meningitis shock areas during
a year declared an epidemic year are 8.2 percentage points (pp) less underweight and 10 pp
less stunted than their non-epidemic year peers. Overall being born in a meningitis shock
district during an epidemic year reduces the current incidence of being underweight by 4.1
pp, versus an increase in the incidence of being underweight of up to 4.1 pp for children born
in meningitis shock districts in years not declared epidemic years. Similarly, being born in
a meningitis shock district during an epidemic year reduces the current incidence of being
stunted by 5.6 pp, versus an increase in the incidence of being stunted of up to 4.4 pp for
children born in meningitis shock districts in years not declared epidemic years.
We find suggestive evidence for the crowd-out of routine vaccination during epidemic
years. While on average, meningitis shocks are associated with increases in total vaccinations,
including routine childhood vaccines for tuberculosis (BCG), polio, diphtheria, pertussis and
tetanus (DPT) and measles, we find heterogeneous effects, depending on whether or not
the WHO declares an epidemic year. Children born in meningitis shock districts during
a declared epidemic year experience a 20% relative reduction in their total vaccinations
received, and their peers born in shock districts during non-epidemic years experience a
13% increase in total vaccinations received relative to the sample mean. While weight and
height improve for children born in meningitis shock areas during declared epidemic years,
routine vaccination falls, as domestic and international organizations focus on meningitis
vaccination in these areas. We conduct a number of robustness checks on our results, and
provide evidence that selective migration does not appear to be a driver of our results.
We show that a primary mechanism explaining the heterogeneity in results and the
reversal of the negative effect of meningitis shocks on economic outcomes during declared
4
epidemic years is the influx of health aid when the WHO announces an epidemic year,
which may offset the negative income shock from increased costs resulting from meningitis
shocks. We document an increase in World Bank health aid projects funded in meningitis
shock districts during declared epidemic years. The funding epidemic effect is redistributive,
with funds flowing away from non-health to health sector projects. The results provide
suggestive evidence that global governance organizations like the WHO play an important
role in mitigating the negative effects of epidemics, partly by coordinating decision making
and funding behavior of international agencies around the disbursement of health aid to
affected regions.
We add to several distinct literatures. First, our work is related to the economics
literature on the economic burden of infectious disease and early life shocks (Acemoglu and
Johnson, 2007; Adhvaryu et al., 2019; Almond, 2006; Bleakley, 2007; Bloom and Mahal, 1997;
Dupas and Robinson, 2013; Adda, 2016; Rangel and Vogl, 2019; McDonald and Roberts,
2006; Maccini and Yang, 2009; Christensen et al., 2021). These studies have demonstrated
that exposure to health shocks like infectious disease in early life can affect various future
life outcomes, including school enrollment, performance and attainment (Bleakley, 2007;
Archibong and Annan, 2017; Fortson, 2011), gender inequality (Archibong and Annan, 2019)
and labor market outcomes (Almond, 2006; Gould, Lavy, and Paserman, 2011; Bhalotra and
Venkataramani, 2015), among others. A more recent literature has focused explicitly on
epidemics and examining the effects of individual and coordinated government responses to
epidemics on societal wellbeing (Fitzpatrick et al., 2021; Maffioli, 2021; Christensen et al.,
2021; Xu, 2021). The studies have highlighted the importance of local accountability in
health systems in managing epidemics like the 2014 Ebola epidemic (Christensen et al.,
2021), and representation in government bureaucracies as a mitigating factor in reducing
mortality during the 1918 pandemic (Xu, 2021). We expand these literatures by providing
quantitative estimates of the economic impacts of epidemics and the role of global governance
5
institutions in mitigating the negative effects of epidemics through policy coordination.
Our work also contributes to the economics literature on the role of aid in development
(Alesina and Dollar, 2000; Burnside and Dollar, 2000; Easterly, 2006; Nunn and Qian, 2014;
Br¨autigam and Knack, 2004; Deserrano, Nansamba, and Qian, 2020; Aldashev, Marini,
and Verdier, 2019). Although a robust literature has found mixed results on the benefits
of foreign aid for development (Burnside and Dollar, 2000; Moyo, 2009), a more recent
literature has noted that health aid may have positive impacts on human capital outcomes,
particularly in asset-constrained regions (Odokonyero et al., 2015; Kotsadam et al., 2018;
Gyimah-Brempong, 2015; Miguel and Kremer, 2004; Bandiera et al., 2019; Ndikumana and
Pickbourn, 2017; Mishra and Newhouse, 2009); and the studies have highlighted heightened
incentives of domestic governments to comply with donor agencies regarding aid in the public
health sector (Dietrich, 2011). Our paper provides quantitative evidence of the positive
effects of health aid in reversing the negative effects of epidemics, where aid increases in
response to epidemic announcements.
The rest of the paper is organized as follows. Section 2 provides a brief background
on the epidemiology and costs of infectious disease, with a focus on meningitis epidemics.
Section 3 describes the data. Section 4 outlines our empirical strategy and presents results
on the effects of meningitis epidemics on human capital development outcomes. Section 5
provides quantitative estimates of the role of World Bank aid as a potential mechanism that
explains the results. Section 6 concludes.
2 Epidemiology and Costs of Epidemics: Evidence from the Menin-
gitis Belt
The WHO defines an epidemic as “the occurrence in a community or region of cases of an
illness clearly in excess of normal expectancy. The number of cases indicating the presence
6
of an epidemic varies according to the agent, size, and type of population exposed, previous
experience or lack of exposure to the disease, and time and place of occurrence.” (WHO,
2020). This definition allows us to distinguish more locally defined disease shocks or ‘local
epidemics’ from officially designated national epidemics by the WHO.
In this paper, we study meningitis epidemics in the African meningitis belt (Figure
1a)4. Meningococcal meningitis is endemic in sub-Saharan Africa. The WHO estimates that
approximately 30,000 cases of the disease are reported annually, with figures rising sharply
in regions during epidemic years5. While cases of meningitis may vary significantly within
a country, national epidemics are declared by the WHO only when the national average
incidence of meningitis is above 100 cases per 100,000 population in a country (de Onis,
2006).
The disease can have high fatality rates of up to 50% when left untreated6. Although
vaccines have been introduced to combat the spread of the disease since the first recorded
cases in 1909 for sub-Saharan Africa, the effectiveness of the vaccines has been limited
because of the mutation and virulence tendencies of the bacterium (LaForce et al., 2009)7.
The periodicity of epidemics in the belt differs by country, with epidemic waves in the
meningitis belt occurring every 8 to 12 years on average by some estimates (Yaka et al.,
2008). Young children and adolescents are especially at risk of infection (Archibong and
Annan, 2017).
4The WHO lists 26 countries in total as being at risk for meningitis epidemics, including Burundi, Rwanda
and Tanzania (WHO, 2018).
5Source: http://www.who.int/mediacentre/factsheets/fs141/en/
6http://www.who.int/mediacentre/factsheets/fs141/en/
7The most recent vaccine MenAfriVac has been available in meningitis belt countries since 2010 and has
been found to be effective against serogroup A, the strain of the bacterium most frequently associated with
epidemics in the belt (Karachaliou et al., 2015). There has been a reduction in serogroup A cases in many
countries since the introduction of the vaccine with the vaccine hailed as a success. Concerns have been
raised about waning herd immunity over the next decade especially if the vaccine does not become part of
routine childhood vaccinations; and an increase in serogroup C cases has been observed in other regions more
recently, prompting concerns about more epidemics from other serogroups of the bacterium (Karachaliou
et al., 2015). Currently, no vaccine prevents all serogroups of Neisseria meningitidis (Yezli et al., 2016).
7
The epidemiology of the disease is complex8. Direct transmission is through contact
with respiratory droplets or throat secretions from infected individuals (LaForce et al., 2009;
Garc´ıa-Pando et al., 2014). The bacteria can be carried in the throat of healthy human
beings, and- for reasons not completely understood- subdue the body’s immune system,
facilitating the spread of infection through the bloodstream to the brain following a 3 to 7
day incubation period (Basta et al., 2018; WHO, 2018)9.
2.1 Costs and Policy Responses to Epidemics
Documented data on health expenditure of countries in the meningitis belt show that house-
holds spend a significant portion of their incomes on direct and indirect costs stemming
from meningitis epidemics (Colombini et al., 2009). In Burkina Faso, a country in the
meningitis belt, households spent around $90 per meningitis case- 34% of per capita GDP-
in direct medical and indirect costs from meningitis infections during the 2006-2007 epidemic
(Colombini et al., 2009). In households affected by sequelae, costs rose to as high as $154
per case. Costs were associated with direct medical expenses from spending on prescriptions
and medicines10 and indirect costs from loss of caregiver income (up to 9 days of lost work),
loss of infected person income (up to 21 days of lost work) and missed school (12 days of
missed school) (Colombini et al., 2009). Meningitis epidemics are a notable negative income
shock to households in the belt.
In its 1998 report on meningococcal meningitis, the WHO recommended a number of
government responses to meningitis epidemics (WHO, 1998). These include the creation of
crisis committees with groups like the Ministry of Health and the WHO to coordinate epi-
8Meningitis epidemics are similar to the COVID-19 pandemic in that they are both spread through
contact with respiratory droplets or throat secretions of infected individuals. An important difference is that
COVID-19 is caused by a virus while meningitis epidemics are caused by a bacterium.
9The WHO estimates that between 10% and 20% of the population carries Neisseria meningitidis in their
throat at any given time, with carriage rate spiking in epidemic years (WHO, 2018).
10Vaccines and treatment are technically free during epidemics, however information asymmetry among
health care workers and shortages of medicines often raise the price of medication (Colombini et al., 2009).
8
demic responses like the dissemination of information to the general public, mass national
vaccination campaigns for the disease and disbursing funds for health projects and antimi-
crobial drugs for treatment11 (WHO, 1998). The costs of full antibacterial therapy treatment
for bacterial meningitis ranged between just under $10 to over $250 (WHO, 1998). The call
to disbursement of funds for health projects and medicines during epidemics is particularly
important for countries in the meningitis belt given that, not only are meningitis epidemics
very costly for households, governments in the meningitis belt spend relatively little on per
capita health spending (Abubakar et al., 2022). By World Bank estimates, as of 2017, gov-
ernment spending was just 23% of total health spending for countries in the meningitis belt,
lower than the within Africa average (35%) and the world average (60%). Additionally, out
of pocket spending as a share of health spending was among the highest in the world at
47% of health spending, compared to 37% within Africa and 18% in the rest of the world.
A key feature of health spending in meningitis belt countries, and in African countries as a
whole, is the high share of health spending coming from external, donor sources. External
spending on health accounts for 23% of health spending in meningitis belt countries, around
the Africa average (20%) and notably higher than the world average (0.2%). Given the
high share of health spending from donor sources and the recommended policy response of
disbursement of funds for health projects during epidemics, international aid has featured
heavily, historically, as a significant part of mitigating the negative effects of epidemics in
the region (Benton and Dionne, 2015).
3 Description of Data: Child Health and World Bank Aid
We combine data from multiple sources for 8 countries in the meningitis belt where data
on meningitis cases and child health outcomes were available, namely, Benin, Burkina Faso,
Cameroon, Ghana, Mali, Nigeria, Niger and Togo (Figure 1b). Further details on the data
11Unlike the COVID-19 pandemic, there were no recommendations for physical distancing or lockdowns
in response to meningitis epidemics.
9
are provided in the proceeding sections and summarized in Table 1.
3.1 Meningitis Cases
We assemble district-level records of mean weekly meningitis cases per 100,000 population
from the WHO from 1986 to 2008 for the 8 countries in the meningitis belt in sub-Saharan
Africa with available data (Figure 1b)12. The WHO data span a number of countries, and
the WHO works with various health ministries to collect survey data from local facilities and
minimize the probability that any errors in measuring health information are systematically
correlated with other outcomes within and across country partners (de Onis, 2006).
As mentioned in Section 2, epidemic years of meningitis are declared by the WHO when
the national average incidence of meningitis is above 100 cases per 100,000 population. Table
1 shows that the average weekly cases per year was around 4 meningitis cases per 100,000 for
the district/years in the full study sample, with significant variability both across and within
countries and years (Figure A1). Following the WHO definition of epidemics as “cases of
an illness clearly in excess of normal expectancy”, we define a “local” epidemic, meningitis
shock, variable, as a measure of the “outside-of-normal expectancy” meningitis events at the
district level. The meningitis shock variable is an indicator that takes on a value equal to one
if meningitis cases in a given year are above the district’s standardized long-term mean. In
other words, the meningitis shock variable equals one if the z-score relative to the district’s
long-term mean of weekly meningitis cases per 100,000 population is greater than zero. We
provide more detail on the specification of the meningitis shock variable in Section 4.
12The WHO data is only available at the district level. District level weekly cases of meningitis case per
100,000 population are available from 1995 to 1999 for 28 districts in Benin, 1996 to 1999 for 30 districts in
Burkina Faso, 1997 to 1998 for 10 districts in Cameroon, 1996 to 1998 for 138 districts in Ghana, 1989 to
1998 for 80 districts in Mali, 1986 to 2008 for 34 districts in Niger, 1995 to 1997 for 116 districts in Nigeria
and 1990 to 1997 for 59 districts in Togo (Figure A1). These make up a dataset of district level meningitis
cases of 495 districts across the 8 countries.
10
3.2 Child Health
To examine the effects of epidemics on child health outcomes, we use geocoded data from the
birth recode (BR) of the Demographic and Health Surveys (DHS) for various years for the
eight study countries. The DHS data are nationally representative cross-sectional household
surveys that provide information on the demographic characteristics of individuals within
households. For the BR sample, women aged 15 to 49 years are individually interviewed
to gather information on every child ever born to the woman. For each of the women
interviewed, the BR has one record for every birth13.
For births within the past 5 years at the time of each survey, the DHS data contains
information on child anthropometric outcomes, including the weight for age z-score (WFA
z) and height for age z-score (HFA z), vaccinations, and mortality status (i.e., whether a
child is alive or dead and age at death if dead). Combined with the district-level meningitis
record, this information provides a dataset of nationally representative individual-level data
of births from 1992 to 2014, covering 14 DHS surveys across the eight study countries14.
The WFA z and HFA z reflect factors that may affect a child’s health in utero, at birth,
and/or after birth. Higher values are generally associated with favorable health conditions
(Jayachandran and Pande, 2017). A child is considered underweight with a WFA z of less
than -2.0; a child is considered stunted with an HFA z of less than -2.0. In the sample,
38% of children are underweight and 36% are stunted. Finally, we examine child vaccination
rates for routine vaccines. We collect available information on BCG (tuberculosis), polio,
DPT (diphtheria, pertussis and tetanus), and measles vaccinations and the total of all vac-
13The BR of the DHS, including important geocoded information on the location of households or house-
hold clusters, is available for 1996, 2001 and 2012 for Benin; 1999, 2003 and 2010 DHS for Burkina Faso;
2004 and 2011 DHS for Cameroon; 1998, 2003, 2008 and 2014 DHS for Ghana; 1996, 2001, 2006 and 2012
DHS for Mali; 1992 and 1998 DHS for Niger; 2003, 2008 and 2013 for Nigeria; and 1998 and 2013 for Togo.
14The final dataset contains data on combined meningitis cases and DHS outcomes for children born
between 1986 and 1999.
11
cinations15. A key feature of these routine vaccines, is that they are offered free of charge
in many low income countries, like the countries in our sample, so the direct costs are often
null, although households may face other indirect costs like insufficient supply or transporta-
tion costs involved in accessing the vaccines (Bobo et al., 2022). 61% and 42% percent of
children in the sample received BCG and measles vaccination, respectively. The average
total number of vaccines received by children in the sample was 3.83 out of a maximum of 8
vaccines (Table 1). Notably, the recommended schedule for routine vaccination of children
by WHO standards is at birth for BCG and with the first dose at birth for polio (Table A1).
The recommendation is near birth for DPT (first dose at 6 weeks). This recommendation
is in contrast with the recommendation for measles, which may be taken much later after
birth (at 9 months) (WHO, 2019).
3.3 World Bank Aid Data
To examine the relationship between WHO epidemic announcements and disaster aid, we use
geocoded data on World Bank funded projects in the International Bank for Reconstruction
and Development (IBRD) and International Development Association (IDA) lending lines by
sectors from AidData (AidData, 2017). This dataset is the only publicly available microlevel
dataset on aid projects for our study region over the years of available data. The World Bank
is a major donor in the Africa region, and was the top donor in Nigeria, the most populous
country on the continent, between 2000 and 2014, funding 31% of recorded aid projects in
the country. The second and third ranked donors in Nigeria were the Bill and Melinda Gates
Foundation (20% of aid projects) and the European Commission (10%) over the same period
(AidData, 2017)16. The World Bank aid data contains the location and sectors of World
Bank funded projects between 1995 and 2014 as shown in Figure 1c. Projects are classified
15There is no information on meningitis vaccination rates in the DHS.
16We explore the effects of epidemic declarations on Official development assistance (ODA) aid from the
Organisation for Economic Co-Operation (OECD) at the country level for 20 countries in the meningitis
belt from 1995 to 2008 in Appendix A.4.
12
by the World Bank as belonging to up to five sectors, such as: health, central government
administration, general public administration, other social services, railways, and roads and
highways. The amount of “aid” or loans and grants (in 2011 USD) committed and disbursed
for each project is also reported. We limit our sample to the subset of projects approved
between 1995 and 2008 to match the duration of our meningitis case data. Summary statistics
in Table 1 show that while on average around $56 million was committed to projects approved
during our study years, only 12% of projects were health projects, where we define a project
as being in the health sector if any one of its five sector categories corresponds to health.
The average duration of these projects was approximately 6 years.
4 Epidemics and Human Capital Development
4.1 Meningitis Shocks and Human Capital Development
We can examine the relationship between meningitis shocks and child health outcomes by
estimating the following equation:
yidctr =αMenin. Shockdct +X0
idctrθ+µd+ηt+ηr+φdr+idctr (1)
where yidctr is the outcome of interest (weight, height and vaccination outcomes in
Section 3) for child iborn in district din country cat time t, whose health outcomes are reg-
istered in survey-year round r. Our main measure of meningitis shocks is “Menin. Shock” or
“Meningitis Shock”, which is an indicator that equals one if a district’s meningitis caseload
z-score, or standard deviation from its long-run average of mean weekly meningitis cases
per 100,000 population, is greater than zero. In practice, we explore two constructions for
“Menin. Shock”: one reflecting the district-level meningitis z-scores, namely, district menin-
gitis positive deviations from their long-run average, and the other capturing the district’s
13
meningitis positive deviations from their long-run (moving) average. We present the results
from the z-score specification in the main text17 , along with results from robustness checks
for marginal changes in the shock measure cutoff, and using the continuous z-score measure.
This specification includes a set of unrestricted within-country district dummies, de-
noted by µd, which capture unobserved differences that are fixed across districts. We also
include year of birth fixed effects to account for potential life cycle changes across cohorts.
Equation 1 includes district-specific trends, φdrthat allow our “Meningitis Shock” and non-
shock districts to follow different trends that may relate to factors like differences in internal
migration patterns that could affect disease transmission18. The child health regressions
also include controls for the mother’s age at birth and level of education, X0
idctr. Errors
are clustered at the district level to allow for arbitrary correlations19 The results are robust
to both specifications, and we present the district level clustering results here. In alternate
specifications and as a robustness check, we estimate Equation 1 using country-by-year fixed
effects to control for aggregate changes that are common across countries over time, e.g.
aggregate prices, and national policies. The results are robust to both specifications, and we
present further inference robustness in Section 4.2.
The intuition behind defining meningitis shock in these ways, as stated previously,
follows the WHO definition of an epidemic, such that an individual district may be ex-
periencing epidemic levels of meningitis cases relative to its expectation, but the national
average does not rise to the level that the WHO declares a country-wide epidemic. A no-
table feature of our shock measure is that there is significant variation in meningitis cases
within country-districts, with no obvious trends in meningitis cases. Districts switch quasi-
randomly between being meningitis shock or non-shock districts from year to year, and there
17The results from both constructions are qualitatively similar, and tables are available upon request.
18This specification is widely used in the health economics literature, following (Maccini and Yang, 2009).
19We estimate all models with standard errors clustered at the district level and Conley standard errors
with a cut-off window of 100 km to account for spatial auto-correlation (Conley, 1999).
14
are no districts that are only shock or only non-shock districts over the years of study. To
test that individuals located in meningitis shock districts an appropriate counterfactual for
those located in unaffected districts, we estimate simple regressions of the likelihood of be-
ing a meningitis shock district, measured as our meningitis shock variable averaged over the
years of available data for each district in each country, on a number of geographic, weather
and institutional characteristics for each district. We show balance across a number of geo-
graphic, weather and institutional characteristics of the meningitis shock measure in Table
A2 in Appendix A.1, and the results in Table A2 show no observable differences in outcomes
across districts that experienced more meningitis shocks between 1986 and 2008 and those
that did not. We interpret the results from Equation 1 as robust associations, and show
that the results are robust to a number of falsification and inference tests, providing further
suggestive evidence of the effects of the meningitis shock measure in Section 4.3.
4.1.1 Mechanisms: The Role of WHO Epidemic Year Announcements
As discussed in Section 2, meningitis epidemics are very costly for households in the African
meningitis belt. While cases of meningitis may vary widely within a country, the WHO
declares a national epidemic only when the number of meningitis cases passes a certain
threshold namely, 100 cases per 100,000 population within the country in a particular year.
Once that happens, primary policy responses recommended by the WHO, as discussed in
Section 2.1, include mass national vaccination campaigns for the disease and disbursing funds
for health projects and antimicrobial drugs for treatment. Countries in the meningitis belt,
and African countries in general, also rely relatively heavily on external donor sources for
health spending as mentioned in Section 2.1. Hence, one hypothesis is that a main channel
that may modulate the effects of meningitis shocks on child health outcomes is through the
WHO announcement of a national epidemic, that could trigger policy responses in the form
of mass vaccination campaigns and disbursement of funds to affected regions. To test this
15
hypothesis, that the effects of meningitis shocks on child health outcomes may be mediated
by WHO announcements of national epidemic years, we estimate Equation 2 below:
yidctr =αMenin. Shockdct +βEpidemic Yearct +γMenin. Shockdct ×Epidemic Yearct
| {z }
+X0
idctrθ+µd+ηt+ηr+φdr+idctr
(2)
where “Epidemic Year” is an indicator that equals one if the WHO declares an epidemic
year in country cand year t, following the aforementioned threshold rule. Our key parameter
of interest γprovides a statistical test of the difference in child health outcomes in meningitis
shock and non-shock districts in WHO announced epidemic years versus non-epidemic years.
This provides an estimate of the “epidemic effect”: how the effects of meningitis epidemics
on human capital development outcomes may be mediated by global policy responses. To
fully explore this “epidemic effect”, we focus on presenting the results from Equation 2 in
this paper, with the results outlined in Section 4.2.
4.2 OLS Estimates
Figure 2 shows the density distributions for two of the main child health outcomes, namely
the weight for age z-scores (WFA z) and height for age z-scores (HFA z) by the meningitis
shock indicator measure. It shows slight stochastic dominance in non-shock districts, where
child WFA z and HFA z are higher in non-shock districts than meningitis shock districts.
These patterns in the raw data are replicated in Figure 3. Figure 3 shows the average share
of stunted and underweight children, along with the average total vaccination received by
children and infant mortality in meningitis shock and non-shock districts for children born
between 1986 and 1999 over the period of available data. Of the 12 years of complete data
available, the share of stunted children is higher in meningitis shock districts in 8 out of 12
years or 67% of the years in the sample. The pattern for the share of underweight children
16
is similar; the share of underweight children is higher in meningitis shock districts in 58%
of the years in the sample. Generally, patterns in the raw data point to more stunted and
more underweight children born in meningitis shock districts versus non-shock districts, in
most years of available data. By contrast, the average total number of vaccinations received
by children is higher in meningitis shock districts in 83% of the years of available data in the
sample. The pattern in the raw data for infant mortality is the least stark of the child health
outcomes, with the share of children born who die within the year higher in meningitis shock
districts in 54% of the years of available data in the sample.
Table 2 presents the OLS estimates of the effects of meningitis shocks on these child
health outcomes, following the specification in Equation 1. The estimates in column (1)
through column (4) of Panel A on child height and weight outcomes, are qualitatively similar
to the depictions in Figure 2 and Figure 3. Meningitis shocks, on average are negatively
associated with child height and weight, and conversely, positively associated with child
stunting and underweight status; the estimates are economically large, though imprecisely
measured. There is no effect of meningitis shocks on infant mortality in column (5) of
Panel A. Columns (1) to (5) of Panel B of Table 2 report estimates for child immunization
outcomes, classified by immunizations recommended at or near birth (BCG, polio, DPT)
versus immunizations recommended much later after birth (measles) as discussed in Section
3. For routine vaccination, the results show significant positive effects of meningitis shocks
on BCG, DPT and the number of polio doses (i.e. at or near birth) vaccinations, with the
signs positive but not significant for measles or non-at/near birth vaccinations. A child born
in a meningitis shock district is significantly more likely to have all her vaccinations than her
peers born in a non-shock district on average, as shown in column (5) of Panel B of Table 2.
The size of the effect is a relative increase of 0.21 vaccinations for children born in meningitis
shock districts- equivalent to a 5.7% increase relative to the sample mean.
17
How do WHO announcements of national epidemic years mediate the effects of menin-
gitis shocks on these child health outcomes? To answer this question, we estimate Equation
2 and examine heterogeneity in the effects of meningitis shocks on child health by WHO
epidemic year declarations. The results are shown in Table 3. Children born in meningitis
shock districts during a declared epidemic year are taller (column (3) of Panel A) and weigh
more (column (1) of Panel A) than their peers born in meningitis shock district during non-
epidemic years. Children born in meningitis shock districts during a declared epidemic year
are 8.2 percentage points (pp) less underweight and 10 pp less stunted than their meningitis
shock, non-epidemic year born peers. Overall being born in a meningitis shock district dur-
ing an epidemic year reduces the current incidence of being underweight by 4.1 pp, versus an
increase in the incidence of being underweight of up to 4.1 pp for children born in meningitis
shock districts in years not declared epidemic years. The total effect is equivalent to a 4.1 pp
and 5.6 pp reduction in the current incidence of being underweight and stunted respectively
for children born in meningitis shock districts in declared epidemic years. The epidemic year
coefficient itself is positive and significant for the child stunting and underweight outcomes,
reflecting the aggregate negative shock effects of epidemics.
Panel B of Table 3 reports estimates for child immunization outcomes. For routine
vaccination, the results show significant negative effects of meningitis shocks in epidemic
years on BCG, DPT and the number of polio doses (i.e. at or near birth) vaccinations, with
the signs negative but not significant for measles or non-at/near birth vaccinations. A child
born in a meningitis shock district during a declared epidemic year is less likely to have all
her vaccinations than her peers born in a meningitis shock district during a non-epidemic
year as shown in column (5) of Panel B. The size of the effect is a relative reduction of
0.72 vaccinations or a total reduction of approximately 0.24 vaccines for children born in
meningitis shock districts during declared epidemic years- equivalent to a 20% and 6.5%
reduction, respectively, relative to the sample mean. The results are reversed for children
18
born in meningitis shock districts during non-epidemic years, who experience an increase in
total vaccinations of up to 0.48 vaccines, or a 13% increase in total vaccinations received
relative to their epidemic year born peers. Our results show that the meningitis shock and
epidemic effects are particularly robust for vaccines that should be administered at or close
to the time of birth20 (Deserrano, Nansamba, and Qian, 2020; Boone, 1996; Br¨autigam and
Knack, 2004).
The results are robust to marginal changes in the meningitis shock indicator cutoff,
above zero (Table 4), and using the continuous z-score measure of meningitis shocks (Table
5). Table 6 shows the results from the model from the alternative specification, replacing
trends with country-by-year fixed effects in Equation 2. The results are robust and the
estimates are largely stable. We provide further inference robustness in Section 4.3 and
discuss possible channels explaining these results in Section 4.4.
4.3 Selective Migration and Inference Robustness
To what extent does migration rationalize our results? We investigate the possibility that
unhealthy individuals (e.g., with low WFA z or low HFA z) might have moved from areas
affected by meningitis to unaffected areas, and as a result, unaffected areas experience low
economic outcomes relative to the affected areas. The dual, though prima facie less plausible,
statement is that more “healthy” individuals might have moved from areas unaffected by
meningitis to the affected areas and as result, unaffected areas experience low economic
outcomes. Thus, instead of assuming limited (selective) internal migration between districts
for identification, we relax this assumption and examine it as an alternative explanation for
our results in Appendix A.2. Although there is no detailed data on internal migration in
our contexts, available evidence suggests that migration is limited in the study region, with
just 9% of the population reporting changing their place of residence between 1988 and 1992
20Thus, we would expect to observe no effects for measles, for example, which should be administered at
9 months.
19
(Bocquier and Traor´e, 1998). In Appendix A.2.2, we conduct trimming exercises, to test that
selective migration is not a driver of the results. Our trimming exercise results in Figure
A2 suggest that migration would have to, differentially, rise by at least 55% to explain the
results, which is very unlikely given the available aforementioned empirical evidence. The
results are consistent with other papers showing a lack of selective migration in developing
country settings (Bazzi et al., 2016).
We conduct a number of inference robustness tests and document that the results are
robust to alternate reporting of standard errors. Inference for our main analyses is based
on the robust cluster (i.e., district) estimator, which allows for arbitrary correlations at the
district-level. We examine the sensitivity of the results to alternative inference procedures by
reporting additional standard errors using (i) robust but unclustered data which adjusts for
arbitrary heteroskedasticity, (ii) the wild-clustered bootstrap, and (iii) the two-way clustering
(i.e., district and time) which accounts for the possibility that errors may be either spatially
or serially correlated. The wild cluster bootstrap is clustered at the district-level and derived
from running 1000 replications in each instance. Overall, the baseline results on inference
show robustness to different procedures with tables available upon request.
We also conduct a number of placebo tests, including an indicator for the epidemic year
2 and 3 years after the child’s year of birth to test the sensitivity of our results to arbitrary
changes in the epidemic year designation. We use the 2 and 3 year leads, and not the 1
year lead, given the positive correlation between the epidemic year indicator in concurrent
year t and the following year t+1 (arising from the fact that some countries may experience
consecutive epidemics). The results show that both the meningitis shock and meningitis
shock x epidemic year interactions are not significant, using the 2 and 3 year leads as shown
in Table 7. In Table 8, we change the epidemic year cutoff, to examine the sensitivity of the
results to changes in the definition of the epidemic year. The meningitis shock x epidemic
20
year interaction, using low numbers (<20 per 100,000 pop.) of meningitis cases nationally
to define the epidemic year, is insignificant with an example shown in Table 8. The results
in Table 7 and Table 8 show no significant effects of erroneous epidemic year designations
on our child health outcomes.
4.4 Channels
What explains the varying results on the average effects of meningitis shocks on child health
outcomes and the heterogeneity in these effects by WHO epidemic year announcements in
Section 4.2? Meningitis shocks are significant negative income shocks for households as
documented in previous literature, and we provide suggestive evidence of reduced economic
activity in meningitis shock districts in Table A4 of Appendix A.3. Lowered household
budgets may leave less income for food and other necessities necessary for proper nutrition,
resulting in malnutrition and reflected in subsequent child stunting and underweight status
in meningitis shock districts.
While there are more stunted and underweight children in meningitis shock districts,
these children receive more vaccination than their peers in non-shock districts. One ex-
planation for this comes from the public health literature on the drivers of routine child
vaccination, particular in lower and lower middle income countries (LIC), like the countries
in our sample (Bobo et al., 2022). As mentioned in Section 3.2, a key feature of these routine
vaccines, is that they are often offered free of charge in LIC countries, so the direct costs
are usually null, although households may face other indirect costs like insufficient supply
or transportation costs involved in accessing the vaccines (Bobo et al., 2022). Often a ma-
jor constraint for child vaccination comes from parental demand for these vaccines, either
because of higher transport costs to access the vaccines or general hesitancy of parents in
these settings to take up these vaccines for their children, due, for example, to past negative
experiences with health institutions (Archibong and Annan, 2021). However, when districts
21
are experiencing a local epidemic, or meningitis shock, poorer households may be more mo-
tivated/willing to seek out free immunization for their children as the least cost medication
available to them, which may explain the higher levels of routine vaccination in meningitis
shock districts.
While there is a paucity of detailed data on household income in the study region, we
can test this hypothesis around the differing income effects of meningitis shocks on child
health outcomes using asset ownership data from the DHS. Using data on assets from the
DHS, we construct a wealth index based on principal components analysis scores and define
liquidity or asset constrained households as those located in the lower parts of the asset
distribution21. Lower wealth households, with wealth in lower than the third quintile of the
wealth distribution, feature largely in the dataset, making up around 43% of the sample on
average. The results in Table 9 provide suggestive evidence that both the negative effects of
meningitis shocks on child height and weight and the positive effects of shocks on vaccination
are concentrated among poorer or more asset-constrained, lower wealth households, in line
with the aforementioned hypotheses.
Meningitis shocks further worsen child height and weight outcomes for children from
lower wealth households, heightening the likelihood of being underweight and stunted by 4.9
pp and 3.8 pp, and increasing the likelihood of infant mortality by 3.8 pp among children in
lower wealth households relative to their lower wealth peers in non-shock districts, as shown
in Panel A of Table 9. The vaccination results in Panel B of Table 9 also show that while
poorer households in non-shock districts are much less likely to vaccinate their children, in
line with previous studies on the links between income and vaccination (Bobo et al., 2022),
poorer households in meningitis shock districts are more likely to vaccinate their children
21Specifically, we conduct a wealth quintile index from 10 assets, with details provided in the Appendix,
and defined asset-constrained or lower wealth households as households situated in lower than the third
quintile of wealth. The analysis assumes stationarity in the wealth status of women’s households and the
results should be interpreted with caution here.
22
in line with our ‘seeking out free/least cost medication during a local epidemic’ hypothesis.
Children in lower wealth households born during meningitis shocks receive 0.21 more total
routine childhood vaccines or 5.4% more total vaccines relative to the sample mean, compared
to their lower wealth peers born in non-shock districts.
What explains the reversal in the effects of meningitis shocks on child health outcomes
during WHO declared national epidemic years, or the epidemic effect? Although there
are multiple possible mechanisms that may explain these reversals, one key channel, comes
from the policy recommendations of the WHO described in Section 2.1, that highlights
responses like mass national vaccination campaigns for the disease and disbursement of
funds for health projects and antimicrobial drugs for treatment during declared epidemic
years. The inflow of health aid to affected regions during declared epidemic years, in a
setting where 23% of health funding comes from external, donor sources, among a largely
poor, asset constrained population, may significantly offset the negative income effects of the
meningitis shock on child weight and height outcomes. By contrast, targeted mass national
vaccination campaigns for meningitis during declared epidemic years may crowd out routine
vaccination either through reduced demand from parents or caregivers as households forego
routine vaccination in favor of the provided meningitis vaccination or through the supply
side, as donor aid focused on epidemic response redirects skilled health workers from routine
immunization to epidemic treatment; a pattern that has been found in numerous settings in
the public health literature (Mansour et al., 2021; Deserrano, Nansamba, and Qian, 2020;
Boone, 1996; Br¨autigam and Knack, 2004; Dinleyici et al., 2021). To test this influx of
health aid hypothesis, we use aid data from World Bank projects described in Section 3.3
and provide further discussion in Section 5.
23
5 Role of Health Aid: Evidence from World Bank Projects
To investigate the inflow of disaster, health aid in response to the epidemic hypothesis, we
use aid data from World Bank projects. As mentioned in Section 3.3, this dataset is the
only publicly available microlevel dataset on aid projects for our study region over the years
of available data; and the World Bank is a major donor in the Africa region, funding 31%
of recorded aid projects in the continent’s most populous country, Nigeria, between 2000
and 2014. We estimate Equation 1 and Equation 2 for individual projects and examine the
effects of meningitis shocks and epidemics on inflows of World Bank aid. We also modify
Equation 2 to specifically test the hypothesis that the World Bank may respond to shocks
during epidemic years with targeted health aid, by examining the triple interaction with an
additional indicator that equals one if the project funded is a health project.
5.1 How World Bank Projects are Approved and Funded
A complete understanding of the results requires some additional information on how World
Bank projects are funded. Research on World Bank internal management practices has been
scant (Ika, Diallo, and Thuillier, 2012); thus we conducted qualitative interviews with World
Bank officials and employees to gain insights into how World Bank aid projects are approved
and funded. Our research revealed that projects take a relatively long time to be approved,
with estimates of an average of 7 to 12 months to approve a single project. Projects must pass
“concept approval, final design approval, then final package to Board” before possibly being
approved and funded. The shortest amount of time to approve projects in an “emergency”
setting is reported to be around 3 to 4 months22 .
What this means is that locations for World Bank health projects are often chosen
ex-ante relative to the declaration of an epidemic year23 (¨
Ohler et al., 2017; Duggan et al.,
22A snapshot of the World Bank project approval process is provided in Figure A3.
23 ¨
Ohler et al. (2017) provide suggestive evidence that projects are targeted geographically by population
share, with more populous regions receiving more projects, rather than by poverty status.
24
2020). This affects the targeting and distribution of health aid because the relatively small
amount of health aid projects funded in the sample (12%) are run by officials who are often
trying to meet particular funding targets in each year. Bank fund managers attempting to
meet certain funding targets for countries may also choose to redirect funds from non-health
projects towards existing health projects in areas with the most need during emergencies
like epidemics. Hence, our results may underestimate the full effect of health aid on human
capital development outcomes during declared epidemic years.
5.2 World Bank Aid Results
Table 10 reports the estimates showing the effects of meningitis shocks on the total amount
of funds (in millions 2011 USD) committed and disbursed to World Bank aid projects. In
column (1) and column (4) of Table 10, there is no significant effect of meningitis shocks on
World Bank aid committed and disbursed respectively. In column (2) and column (5) of the
same table, there is no effect on aid committed and disbursed to meningitis shock districts
during declared epidemic years. In fact, the sign on the epidemic year term is negative and
significant, indicating a reduction of aid to non-shock areas during epidemic years. When we
examine disbursement of aid to health projects in particular, however, the results change.
Column (3) and column (6) of Table 10 show the results for the amount committed and
disbursed to World Bank health aid projects in meningitis shock districts during declared
epidemic years. The triple interaction is positive and significant for both total committed
and total disbursed funds to health projects in meningitis shock districts during epidemic
years in columns (3) and (6).
Meningitis shock districts receive around $52 million more funds in total commitments
to health projects during declared epidemic years over their peers. By contrast, these shock
districts receive less funds to non-health projects during declared epidemic years (-$58 mil-
lion) as shown in column (2). These patterns are replicated for the total amount of funds
25
disbursed by the Bank to shock districts during epidemic years, though the magnitudes are
lower for the amount of funds disbursed to health projects in meningitis shock districts dur-
ing declared epidemic years ($18 million in column (6) of Table 10). There appears to be a
redistribution of funds away from non-health projects and towards health projects in menin-
gitis shock districts during declared epidemic years. There also appears to be a redistribution
of funds committed to health projects during epidemic years away from non-shock districts
towards shock districts as shown in column (3) and column (6). We provide further evidence
of an influx of foreign health aid response to WHO epidemic announcements in Section A.4
in the Appendix.
Although there is sparse publicly available data on the details of the projects approved
over the study period, the dataset includes project titles that provide suggestive evidence
on the kinds of health and non-health projects funded in declared epidemic vs non-epidemic
years24. Notable is the difference between the epidemic and non-epidemic year health project
titles funded. During the epidemic year the top health project titles are ‘health sector and
development program’ and ‘Economic recovery and adjustment credit (ERAC) project’, while
during non-epidemic years, the top project titles are ‘Community action program’, ‘social
fund’ and ‘health, fertility and nutrition project’, providing strong suggestive evidence of the
responsiveness of World bank health funding to epidemic year announcements.
6 Conclusion
Recent scientific literature have provided evidence that future warming may significantly in-
crease the incidence and alter the geographical distribution of aggregate shocks like epidemics
of infectious disease. This may have potentially devastating consequences for global welfare,
absent effective redistributive institutions aimed at improving human capital outcomes.
An important contribution of our paper is to provide quantitative estimates of the
24A snapshot of the top 5 titles in each period is provided in Figure A4.
26
effects of epidemics on human capital development outcomes. We use evidence from the
African meningitis belt, where meningitis is endemic, and examine the effects of meningitis
shocks or local epidemics on human capital development outcomes. We highlight the role
of WHO epidemic year announcements in coordinating policy responses to these shocks
and show heterogeneity in the effects of meningitis shocks by whether or not the WHO
declares an epidemic year. We show that meningitis shocks reduce child health outcomes
on average, increasing the incidence of stunting and underweight status for children born in
shock districts. The effects on reduction in child weight and height are particularly stark for
children born in meningitis shock districts during non-epidemic years. By contrast, children
born in meningitis shock districts during a WHO declared epidemic year are less underweight
and less stunted than their non-epidemic year born peers.
We also document increases in routine child vaccination in meningitis shock districts
on average, where poorer households may seek out free/least cost routine immunization for
their children during periods of meningitis shocks or local epidemics. By contrast, children
born in meningitis shock districts during declared epidemic years receive lower numbers
of routine child vaccinations. We provide suggestive evidence of crowd-out of routine child
vaccination in these shock districts during declared epidemic years, as governance institutions
focus on meningitis vaccination and treatment in these regions, with resulting implications
for both the demand and supply of routine vaccination. We show that a primary mechanism
explaining the reversal in the negative effects of meningitis shocks on child health outcomes
during epidemic years, is an influx of disaster, health aid as a coordinated policy response
when the WHO announces a national epidemic. The results show an increase in World Bank
health aid projects funded in meningitis shock districts during declared epidemic years. The
epidemic funding effect is redistributive, with funds flowing away from non-health projects
towards health sector projects.
27
Our analyses demonstrate that global governance institutions like the WHO play an
important role in mitigating the negative effects of epidemics, partly by coordinating decision
making and funding behavior of international agencies around the disbursement of health
aid to affected regions. Future work will examine the implications of the crowd-out of
routine vaccination effect, particularly in environments of significant vaccine hesitancy in
the aftermath of epidemics (Archibong and Annan, 2021).
a.
b.
c.
Figure 1: Countries in the African Meningitis Belt (a), with districts in study region (b) and
locations of World Bank aid projects for countries and districts in study region over study
years (c)
28
Figure 2: Stochastic dominance: Child weight for age and height for age z-scores are lower
in meningitis shock districts on average
Figure 3: Average child current health outcomes in meningitis shock and non-shock districts.
Share stunted and underweight and total vaccination is higher in most years in meningitis
shock districts.
29
Table 1: Summary Statistics
Statistic N Mean St. Dev. Min Max
District Level Meningitis Data
Meningitis Shock (Indicator) 2,137 0.30 0.46 0.00 1.00
Meningitis Shock (Continuous) 2,137 0.00 0.91 2.00 4.54
Weekly Meningitis Cases (/100,000) 2,282 4.29 11.92 0.00 200.07
Epidemic Year 2,398 0.33 0.47 0.00 1.00
DHS Child Level Data
Infant Mortality 16,486 0.38 0.49 0.00 1.00
WFA z 17,401 1.54 1.33 5.99 5.72
HFA z 17,401 1.47 1.63 6.00 5.89
Underweight 17,401 0.38 0.48 0.00 1.00
Stunted 17,401 0.36 0.48 0.00 1.00
BCG 22,401 0.61 0.49 0.00 1.00
Nos. Polio 22,422 1.45 1.31 0.00 3.00
Nos. DPT 22, 323 1.38 1.34 0.00 3.00
Measles 21, 979 0.42 0.49 0.00 1.00
Nos. Total Vaccines 21,806 3.83 3.33 0.00 8.00
World Bank Project Level Data
Health Project 556 0.12 0.33 0 1
Total Committed, USD 556 55,657,922 28,851,034 5,302,687 238,620,908
Total Disbursed, USD 547 47,585, 463 26,440, 235 1,987, 862 310,653, 294
Project Duration 547 6.117 1.412 1.000 11.000
IEG Outcome 301 3.98 1.24 1.00 6.00
30
Table 2: Effect of meningitis shock on child current weight and height outcomes, at/near
birth (bcg, polio, dpt) versus non-at/near birth recommended (measles) child vaccinations,
total vaccinations and infant mortality
Panel A: Child Health and Infant Mortality Outcomes
Child Weight Child Height Infant Mortality
WFA z Underweight HFA z Stunted Mortality
(1) (2) (3) (4) (5)
Meningitis shock 0.063 0.011 0.075 0.026 0.014
(0.059) (0.020) (0.071) (0.018) (0.011)
Mean of outcome 1.583 0.388 1.476 0.362 0.374
Observations 15,032 15,032 15,032 15,032 15,141
Clusters 135 135 135 135 231
Panel B: Child Vaccination Outcomes
BCG Nos. Polio DPT Measles Total Vaccines
(1) (2) (3) (4) (5)
Meningitis shock 0.026∗∗ 0.0660.0710.027 0.208∗∗
(0.011) (0.038) (0.043) (0.018) (0.099)
Mean of outcome 0.591 1.375 1.328 0.406 3.674
Observations 19,581 19,606 19,548 19,258 19,151
Clusters 136 136 136 136 136
Mother’s controls Yes Yes Yes Yes Yes
District FE Yes Yes Yes Yes Yes
Year of birth FE Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes
Linear time trends (D) Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district. Dependent variables
are child health outcomes described in text from 8 African countries. Mother’s controls include mother’s age at birth and
level of education. Linear time trends (D) are district specific time trends. Year FE are survey year fixed effects. Year of
birth FE and District FE are year of birth fixed effects and district fixed effects respectively. Meningitis shock is the z-score
indicator based on the district-level mean weekly meningitis cases as described in text. ∗∗∗Significant at the 1 percent level,
∗∗Significant at the 5 percent level, Significant at the 10 percent level.
31
Table 3: Effect of meningitis shock on child current weight and height outcomes, at/near
birth (bcg, polio, dpt) versus non-at/near birth recommended (measles) child vaccinations,
total vaccinations and infant mortality in epidemic versus non-epidemic years
Panel A: Child Health and Infant Mortality Outcomes
Child Weight Child Height Infant Mortality
WFA z Underweight HFA z Stunted Mortality
(1) (2) (3) (4) (5)
Meningitis shock 0.209∗∗ 0.041 0.1560.044∗∗ 0.009
(0.094) (0.027) (0.082) (0.020) (0.018)
Epidemic year 0.245∗∗ 0.072∗∗ 0.614∗∗∗ 0.174∗∗∗ 0.058∗∗∗
(0.103) (0.036) (0.119) (0.034) (0.020)
Meningitis shock
x Epidemic year 0.353∗∗ 0.0820.388∗∗∗ 0.100∗∗∗ 0.010
(0.139) (0.042) (0.124) (0.036) (0.020)
Mean of outcome 1.583 0.388 1.476 0.362 0.374
Observations 15,032 15,032 15,032 15,032 15,141
Clusters 135 135 135 135 231
Panel B: Child Vaccination Outcomes
BCG Nos. Polio DPT Measles Total Vaccines
(1) (2) (3) (4) (5)
Meningitis shock 0.065∗∗∗ 0.183∗∗∗ 0.174∗∗∗ 0.035 0.476∗∗∗
(0.016) (0.062) (0.064) (0.033) (0.160)
Epidemic year 0.053∗∗ 0.194∗∗ 0.1750.131∗∗∗ 0.549∗∗
(0.022) (0.076) (0.091) (0.042) (0.216)
Meningitis shock
x Epidemic year 0.092∗∗∗ 0.293∗∗∗ 0.259∗∗ 0.067 0.719∗∗∗
(0.026) (0.095) (0.108) (0.052) (0.265)
Mean of outcome 0.591 1.375 1.328 0.406 3.674
Observations 19,581 19,606 19,548 19,258 19,151
Clusters 136 136 136 136 136
Mother’s controls Yes Yes Yes Yes Yes
District FE Yes Yes Yes Yes Yes
Year of birth FE Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes
Linear time trends (D) Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district. Dependent variables are child health
outcomes described in text from 8 African countries. Mother’s controls include mother’s age at birth and level of education. Linear time
trends (D) are district specific time trends. Year FE are survey year fixed effects. Year of birth FE and District FE are year of birth fixed
effects and district fixed effects respectively. Meningitis sho ck is the z-score indicator based on the district-level mean weekly meningitis
cases as described in text. Epidemic year is an indicator that equals one if the WHO declares an epidemic year. ∗∗∗Significant at the 1
percent level, ∗∗Significant at the 5 percent level, Significant at the 10 percent level.
32
Table 4: Robustness to marginal changes in shock cutoff: Effect of meningitis shock (>
0.5) on child current weight and height outcomes, at/near birth (bcg, polio, dpt) versus
non-at/near birth recommended (measles) child vaccinations, total vaccinations and infant
mortality in epidemic versus non-epidemic years
Panel A: Child Health and Infant Mortality Outcomes
Child Weight Child Height Infant Mortality
WFA z Underweight HFA z Stunted Mortality
(1) (2) (3) (4) (5)
Meningitis shock 0.253∗∗ 0.054 0.1920.052∗∗ 0.009
(0.125) (0.034) (0.107) (0.025) (0.025)
Epidemic year 0.227∗∗∗ 0.072∗∗ 0.571∗∗∗ 0.161∗∗∗ 0.052∗∗∗
(0.082) (0.029) (0.111) (0.032) (0.020)
Meningitis shock
x Epidemic year 0.486∗∗∗ 0.123∗∗∗ 0.480∗∗∗ 0.119∗∗∗ 0.016
(0.153) (0.045) (0.147) (0.037) (0.028)
Mean of outcome 1.583 0.388 1.476 0.362 0.374
Observations 15,032 15,032 15,032 15,032 15,141
Clusters 135 135 135 135 231
Panel B: Child Vaccination Outcomes
BCG Nos. Polio DPT Measles Total Vaccines
(1) (2) (3) (4) (5)
Meningitis shock 0.064∗∗∗ 0.1310.114 0.029 0.353
(0.021) (0.079) (0.076) (0.034) (0.191)
Epidemic year 0.0340.145∗∗ 0.1450.126∗∗∗ 0.447∗∗
(0.021) (0.067) (0.080) (0.039) (0.189)
Meningitis shock
x Epidemic year 0.087∗∗∗ 0.247∗∗ 0.230∗∗ 0.074 0.651∗∗
(0.029) (0.109) (0.112) (0.046) (0.278)
Mean of outcome 0.591 1.375 1.328 0.406 3.674
Observations 19,581 19,606 19,548 19,258 19,151
Clusters 136 136 136 136 136
Mother’s controls Yes Yes Yes Yes Yes
District FE Yes Yes Yes Yes Yes
Year of birth FE Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes
Linear time trends (D) Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district. Dependent variables are child health
outcomes described in text from 8 African countries. Mother’s controls include mother’s age at birth and level of education. Linear time
trends (D) are district specific time trends. Year FE are survey year fixed effects. Year of birth FE and District FE are year of birth fixed
effects and district fixed effects respectively. Meningitis sho ck is the z-score indicator based on the district-level mean weekly meningitis
cases as described in text. Epidemic year is an indicator that equals one if the WHO declares an epidemic year. ∗∗∗Significant at the 1
percent level, ∗∗Significant at the 5 percent level, Significant at the 10 percent level.
33
Table 5: Robustness to continuous measure: Effect of meningitis shock (continuous z-score)
on child current weight and height outcomes, at/near birth (bcg, polio, dpt) versus non-
at/near birth recommended (measles) child vaccinations, total vaccinations and infant mor-
tality in epidemic versus non-epidemic years
Panel A: Child Health and Infant Mortality Outcomes
Child Weight Child Height Infant Mortality
WFA z Underweight HFA z Stunted Mortality
(1) (2) (3) (4) (5)
Meningitis shock (C) 0.214∗∗∗ 0.043∗∗ 0.148∗∗ 0.039∗∗ 0.008
(0.072) (0.021) (0.066) (0.017) (0.016)
Epidemic year 0.1420.058∗∗ 0.519∗∗∗ 0.146∗∗∗ 0.053∗∗
(0.084) (0.027) (0.098) (0.029) (0.020)
Meningitis shock (C)
x Epidemic year 0.325∗∗∗ 0.080∗∗∗ 0.301∗∗∗ 0.073∗∗∗ 0.007
(0.082) (0.024) (0.084) (0.021) (0.016)
Mean of outcome 1.583 0.388 1.476 0.362 0.374
Observations 15,032 15,032 15,032 15,032 15,141
Clusters 135 135 135 135 231
Panel B: Child Vaccination Outcomes
BCG Nos. Polio DPT Measles Total Vaccines
(1) (2) (3) (4) (5)
Meningitis shock (C) 0.049∗∗∗ 0.129∗∗ 0.122∗∗ 0.022 0.335∗∗∗
(0.012) (0.050) (0.048) (0.020) (0.121)
Epidemic year 0.016 0.093 0.095 0.117∗∗∗ 0.313
(0.018) (0.067) (0.078) (0.039) (0.184)
Meningitis shock (C)
x Epidemic year 0.060∗∗∗ 0.184∗∗∗ 0.178∗∗∗ 0.0450.479∗∗∗
(0.015) (0.059) (0.058) (0.023) (0.147)
Mean of outcome 0.591 1.375 1.328 0.406 3.674
Observations 19,581 19,606 19,548 19,258 19,151
Clusters 136 136 136 136 136
Mother’s controls Yes Yes Yes Yes Yes
District FE Yes Yes Yes Yes Yes
Year of birth FE Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes
Linear time trends (D) Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district. Dependent variables are child health
outcomes described in text from 8 African countries. Mother’s controls include mother’s age at birth and level of education. Linear time
trends (D) are district specific time trends. Year FE are survey year fixed effects. Year of birth FE and District FE are year of birth
fixed effects and district fixed effects respectively. Meningitis shock (C) is the continuous z-score based on the district-level mean weekly
meningitis cases as described in text. Epidemic year is an indicator that equals one if the WHO declares an epidemic year. ∗∗∗Significant
at the 1 percent level, ∗∗Significant at the 5 percent level, Significant at the 10 percent level.
34
Table 6: Robustness: Effect of meningitis shock on child current weight and height outcomes,
at/near birth (bcg, polio, dpt) versus non-at/near birth recommended (measles) child vac-
cinations, total vaccinations and infant mortality in epidemic versus non-epidemic years
Panel A: Child Health and Infant Mortality Outcomes
Child Weight Child Height Infant Mortality
WFA z Underweight HFA z Stunted Mortality
(1) (2) (3) (4) (5)
Meningitis shock 0.176∗∗ 0.0430.178∗∗ 0.046∗∗ 0.009
(0.081) (0.022) (0.079) (0.020) (0.018)
Epidemic year 0.217∗∗ 0.0610.543∗∗∗ 0.157∗∗∗ 0.055∗∗∗
(0.101) (0.036) (0.126) (0.037) (0.019)
Meningitis shock
x Epidemic year 0.288∗∗ 0.0660.301∗∗∗ 0.076∗∗ 0.009
(0.119) (0.036) (0.114) (0.034) (0.021)
Mean of outcome 1.583 0.388 1.476 0.362 0.374
Observations 15,032 15,032 15,032 15,032 15,141
Clusters 135 135 135 135 231
Panel B: Child Vaccination Outcomes
BCG Nos. Polio DPT Measles Total Vaccines
(1) (2) (3) (4) (5)
Meningitis shock 0.047∗∗ 0.120∗∗ 0.1180.014 0.309
(0.018) (0.060) (0.061) (0.037) (0.166)
Epidemic year 0.048∗∗ 0.164∗∗ 0.145 0.113∗∗∗ 0.461∗∗
(0.022) (0.075) (0.091) (0.041) (0.216)
Meningitis shock
x Epidemic year 0.068∗∗ 0.198∗∗ 0.167 0.024 0.454
(0.027) (0.094) (0.103) (0.052) (0.261)
Mean of outcome 0.591 1.375 1.328 0.406 3.674
Observations 19,581 19,606 19,548 19,258 19,151
Clusters 136 136 136 136 136
Mother’s controls Yes Yes Yes Yes Yes
District FE Yes Yes Yes Yes Yes
Year of birth FE Yes Yes Yes Yes Yes
Country x year FE Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district. Dependent variables are child
health outcomes described in text from 8 African countries. Mother’s controls include mother’s age at birth and level of education.
Country x year fixed effects (FE) are country x survey year FE. Year of birth FE and District FE are year of birth fixed effects
and district fixed effects respectively. Meningitis shock is the z-score indicator based on the district-level mean weekly meningitis
cases as described in text. Epidemic year is an indicator that equals one if the WHO declares an epidemic year. ∗∗∗ Significant at
the 1 percent level, ∗∗Significant at the 5 percent level, Significant at the 10 percent level.
35
Table 7: Placebo tests for Epidemic year: Effect of meningitis shock on child health (weight-
for-age WFA zand height-for-age HFA z)
Child Weight Child Height
WFA z HFA z
(1) (2) (3) (4) (5) (6)
Meningitis shock 0.209∗∗ 0.051 0.041 0.1560.008 0.063
(0.094) (0.067) (0.064) (0.091) (0.083) (0.078)
Epidemic year 0.245∗∗ 0.614∗∗∗
(0.103) (0.119)
Epidemic year, t+2 0.248 0.510∗∗∗
(0.154) (0.159)
Epidemic year, t+3 0.730∗∗∗ 0.764∗∗∗
(0.207) (0.210)
Meningitis shock
x Epidemic year 0.353∗∗ 0.388∗∗∗
(0.139) (0.124)
Meningitis shock
x Epidemic year t+2 0.179 0.157
(0.130) (0.179)
Meningitis shock
x Epidemic year t+3 0.107 0.078
(0.165) (0.181)
Mean of outcome 1.583 1.583 1.583 1.476 1.476 1.476
Observations 15,032 15,032 15,032 15,032 15,032 15,032
Clusters 135 135 135 135 135 135
Mother’s controls Yes Yes Yes Yes Yes Yes
District FE Yes Yes Yes Yes Yes Yes
Year of birth FE Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Linear time trends (D) Yes Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district. Dependent variables are child health outcomes described
in text from 8 African countries and are WFA z scores in columns (1) to (3), and HFA z scores in columns (4) to (6). Epidemic year t+2 is an indicator
for an epidemic year 2 years after the child’s year of birth; t+3 is 3 years after the child’s year of birth. Mother’s controls include mother’s age at birth
and level of education. Linear time trends (D) are district specific time trends. Year FE are survey year fixed effects. Meningitis sho ck is z-score indicator
based on district-level mean and equal to 1 if the z-score is greater than 0. ∗∗∗Significant at the 1 percent level, ∗∗Significant at the 5 percent level,
Significant at the 10 percent level.
36
Table 8: Placebo tests for Epidemic year: Changing epidemic year cutoff
Child Weight Child Height
WFA z Underweight HFA z Stunted
(1) (2) (3) (4)
Meningitis shock 0.197 0.053 0.000 0.050
(0.172) (0.067) (0.159) (0.056)
Epidemic year (>5) 0.214 0.099 0.513∗∗∗ 0.042
(0.196) (0.075) (0.184) (0.061)
Meningitis shock
x Epidemic year (>5) 0.147 0.049 0.044 0.071
(0.205) (0.077) (0.197) (0.065)
Mean of outcome 1.583 0.388 1.476 0.362
Observations 15,032 15,032 15,032 15,032
Clusters 135 135 135 135
Mother’s controls Yes Yes Yes Yes
District FE Yes Yes Yes Yes
Year of birth FE Yes Yes Yes Yes
Year FE Yes Yes Yes Yes
Linear time trends (D) Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district. Epidemic
year is an indicator that equals 1 if the number of meningitis cases is greater than 5 per 100,000 population
in the year. Dependent variables are child health outcomes described in text from 8 African countries.
Mother’s controls include mother’s age at birth and level of education. Linear time trends (D) are district
specific time trends. Year FE are survey year fixed effects. Meningitis shock is z-score indicator based on
district-level mean as described in text. ∗∗∗Significant at the 1 percent level, ∗∗Significant at the 5 percent
level, Significant at the 10 percent level.
37
Table 9: Effect of meningitis shock on child current weight and height outcomes, at/near
birth (bcg, polio, dpt) versus non-at/near birth recommended (measles) child vaccinations,
total vaccinations and infant mortality by wealth status
Panel A: Child Health and Infant Mortality Outcomes
Child Weight Child Height Infant Mortality
WFA z Underweight HFA z Stunted Mortality
(1) (2) (3) (4) (5)
Meningitis shock 0.015 0.015 0.009 0.007 0.034∗∗
(0.072) (0.025) (0.091) (0.023) (0.017)
Lower wealth 0.180∗∗∗ 0.071∗∗∗ 0.234∗∗∗ 0.064∗∗∗ 0.027∗∗
(0.046) (0.016) (0.057) (0.014) (0.012)
Meningitis shock
x Lower wealth 0.151∗∗ 0.049∗∗ 0.1610.038 0.038
(0.062) (0.023) (0.084) (0.025) (0.020)
Mean of outcome 1.583 0.388 1.476 0.362 0.374
Observations 15,032 15,032 15,032 15,032 15,141
Clusters 135 135 135 135 231
Panel B: Child Vaccination Outcomes
BCG Nos. Polio DPT Measles Total Vaccines
(1) (2) (3) (4) (5)
Meningitis shock 0.005 0.025 0.032 0.024 0.099
(0.014) (0.044) (0.051) (0.021) (0.115)
Lower wealth 0.195∗∗∗ 0.452∗∗∗ 0.490∗∗∗ 0.125∗∗∗ 1.284∗∗∗
(0.024) (0.055) (0.057) (0.015) (0.015)
Meningitis shock
x Lower wealth 0.059∗∗∗ 0.0790.077 0.007 0.210
(0.018) (0.043) (0.049) (0.021) (0.117)
Mean of outcome 0.591 1.375 1.328 0.406 3.674
Observations 19,581 19,606 19,548 19,258 19,151
Clusters 136 136 136 136 136
Mother’s controls Yes Yes Yes Yes Yes
District FE Yes Yes Yes Yes Yes
Year of birth FE Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes
Linear time trends (D) Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district. Dependent variables are child health
outcomes described in text from 8 African countries. Mother’s controls include mother’s age at birth and level of education. Linear time
trends (D) are district specific time trends. Year FE are survey year fixed effects. Year of birth FE and District FE are year of birth fixed
effects and district fixed effects respectively. Meningitis sho ck is the z-score indicator based on the district-level mean weekly meningitis
cases as described in text. Lower wealth is an indicator that equals one if the household is in less than the third quintile for wealth in
the sample based on the wealth index calculated from the DHS using principal component analysis of asset ownership, as described in
text. The wealth index is a 1 to 5 categorical variable where 1 is the poorest quintile and 5 is the richest quintile. So Lower wealth is an
indicator that equals one if the household wealth index is less than 3. ∗∗∗Significant at the 1 percent level, ∗∗Significant at the 5 percent
level, Significant at the 10 percent level.
38
References
Abubakar, Ibrahim, Sarah L Dalglish, Blake Angell, Olutobi Sanuade, Seye Abimbola,
Aishatu Lawal Adamu, Ifedayo MO Adetifa, Tim Colbourn, Afolabi Olaniyi Ogunlesi,
Obinna Onwujekwe et al. 2022. “The Lancet Nigeria Commission: Investing in health and
the future of the nation.The Lancet 399 (10330): 1155–1200.
Acemoglu, Daron, and Simon Johnson. 2007. “Disease and development: the effect of life
expectancy on economic growth.Journal of Political Economy 115 (6): 925–985.
Adda, J´erˆome. 2016. “Economic activity and the spread of viral diseases: Evidence from
high frequency data.The Quarterly Journal of Economics 131 (2): 891–941.
Adhvaryu, Achyuta, Prashant Bharadwaj, James Fenske, Anant Nyshadham, Richard Stan-
ley et al. 2019. “Dust and Death: Evidence from the West African Harmattan.Forth-
coming in The Economic Journal .
AidData. 2017. “WorldBank-GeocodedResearchRelease-Level1-v1.4.2 geocoded dataset.
AidData.Williamsburg, VA and Washington, DC: AidData.http://aiddata.org/
research-datasets.
Aldashev, Gani, Marco Marini, and Thierry Verdier. 2019. “Samaritan Bundles: Fundraising
Competition and Inefficient Clustering in NGO Projects.The Economic Journal .
Alesina, Alberto, and David Dollar. 2000. “Who gives foreign aid to whom and why?”
Journal of economic growth 5 (1): 33–63.
Almond, Douglas. 2006. “Is the 1918 Influenza pandemic over? Long-term effects of in utero
Influenza exposure in the post-1940 US population.Journal of Political Economy 114 (4):
672–712.
40
Archibong, Belinda. 2019. “Explaining divergence in the long-term effects of precolonial
centralization on access to public infrastructure services in Nigeria.World Development
121: 123–140.
Archibong, Belinda, and Francis Annan. 2017. “Disease and Gender Gaps in Human Cap-
ital Investment: Evidence from Niger’s 1986 Meningitis Epidemic.American Economic
Review, Papers and Proceedings 107 (5): 530–35.
Archibong, Belinda, and Francis Annan. 2019. “Schooling in Sickness and in Health: The
Effects of Epidemic Disease on Gender Inequality.CDEP-CGEG Working Paper No. 54 .
Archibong, Belinda, and Francis Annan. 2021. ‘We Are Not Guinea Pigs’: The Effects of
Negative News on Vaccine Compliance.Brookings Brooke Shearer Series Number 10 .
Bandiera, Oriana, Niklas Buehren, Markus P Goldstein, Imran Rasul, and Andrea Smurra.
2019. The Economic Lives of Young Women in the Time of Ebola: Lessons from an
Empowerment Program. The World Bank.
Basta, Nicole E, Abdoulaye Berthe, Mahamadou Keita, Uma Onwuchekwa, Boubou Tam-
boura, Awa Traore, Musa Hassan-King, Olivier Manigart, Maria Nascimento, James M
Stuart et al. 2018. “Meningococcal carriage within households in the African meningitis
belt: A longitudinal pilot study.Journal of Infection 76 (2): 140–148.
Bazzi, Samuel, Arya Gaduh, Alexander D Rothenberg, and Maisy Wong. 2016. “Skill trans-
ferability, migration, and development: Evidence from population resettlement in Indone-
sia.American Economic Review 106 (9): 2658–98.
Beauchemin, Cris, and Philippe Bocquier. 2004. “Migration and urbanisation in Francophone
West Africa: An overview of the recent empirical evidence.Urban Studies 41 (11): 2245–
2272.
41
Bell, Martin, Elin Charles-Edwards, Philipp Ueffing, John Stillwell, Marek Kupiszewski, and
Dorota Kupiszewska. 2015. “Internal migration and development: comparing migration
intensities around the world.Population and Development Review 41 (1): 33–58.
Benton, Adia, and Kim Yi Dionne. 2015. “International political economy and the 2014
West African Ebola outbreak.African Studies Review 58 (1): 223–236.
Bernard, Aude, and Martin Bell. 2018. “Internal migration and education: A cross-national
comparison.arXiv preprint arXiv:1812.08913 .
Bhalotra, Sonia R, and Atheendar Venkataramani. 2015. “Shadows of the captain of the men
of death: Early life health interventions, human capital investments, and institutions..
Bisset, S, M Fournier, L Pagani, and M Janosz. 2013. “Predicting academic and cogni-
tive outcomes from weight status trajectories during childhood.International journal of
obesity 37 (1): 154–159.
Bleakley, Hoyt. 2007. “Disease and development: evidence from hookworm eradication in
the American South.The Quarterly Journal of Economics 122 (1): 73–117.
Bloom, David E, and Ajay S Mahal. 1997. “Does the AIDS epidemic threaten economic
growth?” Journal of Econometrics 77 (1): 105–124.
Bobo, Firew Tekle, Augustine Asante, Mirkuzie Woldie, Angela Dawson, and Andrew Hayen.
2022. “Child vaccination in sub-Saharan Africa: Increasing coverage addresses inequali-
ties.Vaccine 40 (1): 141–150.
Bocquier, Philippe, and Sadio Traor´e. 1998. “Migration and urbanization in West Africa:
Methodological issues in data collection and inference.Migration, Urbanization, and De-
velopment: New Directions and Issues. New York: United Nations Population Fund and
Kluwer Academic Publishers pp. 249–273.
42
Boone, Peter. 1996. “Politics and the effectiveness of foreign aid.European Economic Review
40 (2): 289–329.
Br¨autigam, Deborah A, and Stephen Knack. 2004. “Foreign aid, institutions, and governance
in sub-Saharan Africa.Economic development and cultural change 52 (2): 255–285.
Burnside, Craig, and David Dollar. 2000. “Aid, policies, and growth.American economic
review 90 (4): 847–868.
Christensen, Darin, Oeindrila Dube, Johannes Haushofer, Bilal Siddiqi, and Maarten Voors.
2021. “Building resilient health systems: experimental evidence from Sierra Leone and
the 2014 Ebola outbreak.The Quarterly Journal of Economics 136 (2): 1145–1198.
Colombini, Ana¨ıs, Fernand Bationo, Sylvie Zongo, Fatoumata Ouattara, Ousmane Badolo,
Philippe Jaillard, Emmanuel Seini, Bradford D Gessner, and Alfred Da Silva. 2009. “Costs
for households and community perception of meningitis epidemics in Burkina Faso.Clin-
ical Infectious Diseases 49 (10): 1520–1525.
Conley, Timothy G. 1999. “GMM estimation with cross sectional dependence.Journal of
econometrics 92 (1): 1–45.
de Onis, Mercedes. 2006. “Reliability of anthropometric measurements in the WHO Multi-
centre Growth Reference Study.Acta Paediatrica 95: 38–46.
Deserrano, Erika, Aisha Nansamba, and Nancy Qian. 2020. “The unintended consequences
of NGO-provided aid on government services in Uganda.NBER Working Paper .
Dietrich, Simone. 2011. “The politics of public health aid: why corrupt governments have
incentives to implement aid effectively.World Development 39 (1): 55–63.
43
Dinleyici, Ener Cagri, Ray Borrow, Marco Aur´elio Palazzi Safadi, Pierre van Damme, and
Flor M Munoz. 2021. “Vaccines and routine immunization strategies during the COVID-19
pandemic.Human vaccines & immunotherapeutics 17 (2): 400–407.
Duggan, Julian, Scott Morris, Justin Sandefur, and George Yang. 2020. “Is the World Bank’s
COVID-19 crisis lending big enough, fast enough? New evidence on loan disbursements.
Center for Global Development Working Paper 554 .
Dupas, Pascaline, and Jonathan Robinson. 2013. “Why don’t the poor save more? Evidence
from health savings experiments.American Economic Review 103 (4): 1138–71.
Easterly, William. 2006. The white man’s burden: why the West’s efforts to aid the rest have
done so much ill and so little good. Penguin Press New York.
Fitzpatrick, Anne E, Sabrin A Beg, Laura C Derksen, Anne Karing, Jason T Kerwin, Adri-
enne Lucas, Natalia Ordaz Reynoso, and Munir Squires. 2021. “Health Knowledge and
Non-Pharmaceutical Interventions During the COVID-19 Pandemic in Africa.Journal of
Economic Behavior & Organization 190: 33–53.
Fortson, Jane G. 2011. “Mortality risk and human capital investment: The Impact of
HIV/AIDS in Sub-Saharan Africa.The Review of Economics and Statistics 93 (1): 1–15.
Garc´ıa-Pando, Carlos P´erez, Madeleine C Thomson, Michelle C Stanton, Peter J Diggle,
Thomas Hopson, Rajul Pandya, Ron L Miller, and St´ephane Hugonnet. 2014. “Meningitis
and climate: from science to practice.Earth Perspectives 1 (1): 14.
Gould, Eric D, Victor Lavy, and M Daniele Paserman. 2011. “Sixty years after the magic
carpet ride: The long-run effect of the early childhood environment on social and economic
outcomes.The Review of Economic Studies 78 (3): 938–973.
44
Gyimah-Brempong, Kwabena. 2015. “Do African countries get health from health aid?”
Journal of African Development 17 (2): 83–114.
Henderson, Vernon, Adam Storeygard, and David N Weil. 2011. “A bright idea for measuring
economic growth.American Economic Review 101 (3): 194–99.
Ika, Lavagnon A, Amadou Diallo, and Denis Thuillier. 2012. “Critical success factors for
World Bank projects: An empirical investigation.International journal of project man-
agement 30 (1): 105–116.
Jayachandran, Seema, and Rohini Pande. 2017. “Why are Indian children so short? The
role of birth order and son preference.American Economic Review 107 (9): 2600–2629.
Karachaliou, Andromachi, Andrew JK Conlan, Marie-Pierre Preziosi, and Caroline L Trot-
ter. 2015. “Modeling long-term vaccination strategies with MenAfriVac in the African
meningitis belt. Clinical Infectious Diseases 61 (suppl 5): S594–S600.
Kiszewski, Anthony, Andrew Mellinger, Andrew Spielman, Pia Malaney, Sonia Ehrlich Sachs,
and Jeffrey Sachs. 2004. “A global index representing the stability of malaria transmission.
The American journal of tropical medicine and hygiene 70 (5): 486–498.
Kotsadam, Andreas, Gudrun Østby, Siri Aas Rustad, Andreas Forø Tollefsen, and Henrik
Urdal. 2018. “Development aid and infant mortality. Micro-level evidence from Nigeria.
World Development 105: 59–69.
LaForce, F Marc, Neil Ravenscroft, Mamoudou Djingarey, and Simonetta Viviani. 2009.
“Epidemic meningitis due to Group A Neisseria meningitidis in the African meningitis
belt.Vaccine 27: B13–B19.
Maccini, Sharon, and Dean Yang. 2009. “Under the weather: Health, schooling, and eco-
nomic consequences of early-life rainfall.American Economic Review 99 (3): 1006–26.
45
Maffioli, Elisa M. 2021. “The political economy of health epidemics: evidence from the Ebola
outbreak.Journal of Development Economics 151: 102651.
Mansour, Ziad, Jinan Arab, Racha Said, Alissar Rady, Randa Hamadeh, Bernard Gerbaka,
and Abdul Rahman Bizri. 2021. “Impact of COVID-19 pandemic on the utilization of
routine immunization services in Lebanon.Plos one 16 (2): e0246951.
McDonald, Scott, and Jennifer Roberts. 2006. “AIDS and economic growth: A human
capital approach.Journal of Development Economics 80 (1): 228–250.
Michalopoulos, Stelios, and Elias Papaioannou. 2013. “Pre-colonial ethnic institutions and
contemporary African development.Econometrica 81 (1): 113–152.
Miguel, Edward, and Michael Kremer. 2004. “Worms: identifying impacts on education and
health in the presence of treatment externalities.Econometrica 72 (1): 159–217.
Mishra, Prachi, and David Newhouse. 2009. “Does health aid matter?” Journal of health
economics 28 (4): 855–872.
Moyo, Dambisa. 2009. Dead aid: Why aid is not working and how there is a better way for
Africa. Macmillan.
Murdock, George Peter. 1967. “Ethnographic atlas: a summary.Ethnology 6 (2): 109–236.
Ndikumana, L´eonce, and Lynda Pickbourn. 2017. “The impact of foreign aid allocation on
access to social services in sub-Saharan Africa: The case of water and sanitation.World
Development 90: 104–114.
Nunn, Nathan, and Nancy Qian. 2014. “US food aid and civil conflict.American Economic
Review 104 (6): 1630–66.
46
Odokonyero, Tonny, Alex Ijjo, Robert Marty, Tony Muhumuza, and Godfrey Owot Moses.
2015. Subnational Perspectives on Aid Effectiveness: Impact of Aid on Health Outcomes
in Uganda. Technical report.
¨
Ohler, Hannes, Mario Negre, Lodewijk Smets, Renzo Massari, and ˇ
Zeljko Bogeti´c. 2017.
“Putting your money where your mouth is: geographic targeting of World Bank projects
to the bottom 40 percent.World Bank Policy Research Working Paper .
Rangel, Marcos A, and Tom S Vogl. 2019. “Agricultural fires and health at birth.Review
of Economics and Statistics 101 (4): 616–630.
WHO. 1998. “Control of epidemic meningococcal disease: WHO practical guidelines..
WHO. 2018. “Meningococcal meningitis. Fact sheet 2018..
WHO. 2019. “WHO recommendations for routine immunizations summary tables..
WHO. 2020. “WHO Humanitarian Health Action, Definitions: emergencies.https://www.
who.int/hac/about/definitions/en/.
Xu, Guo. 2021. “Bureaucratic Representation and State Responsiveness during Times of
Crisis: The 1918 Pandemic in India.The Review of Economics and Statistics pp. 1–29.
Yaka, Pascal, Benjamin Sultan, H´el`ene Broutin, Serge Janicot, Solenne Philippon, and Nicole
Fourquet. 2008. “Relationships between climate and year-to-year variability in meningitis
outbreaks: a case study in Burkina Faso and Niger.International Journal of Health
Geographics 7 (1): 34.
Yezli, Saber, Abdulaziz A Bin Saeed, Abdullah M Assiri, Rafat F Alhakeem, Muslim A
Yunus, Abdulhafiz M Turkistani, Robert Booy, and Badriah M Alotaibi. 2016. “Prevention
of meningococcal disease during the Hajj and Umrah mass gatherings: past and current
measures and future prospects.International Journal of Infectious Diseases 47: 71–78.
47
A Appendix (For Online Publication)
Contents
1 Introduction 2
2 Epidemiology and Costs of Epidemics: Evidence from the Meningitis Belt 6
2.1 Costs and Policy Responses to Epidemics . . . . . . . . . . . . . . . . . . . . 8
3 Description of Data: Child Health and World Bank Aid 9
3.1 MeningitisCases ................................. 10
3.2 ChildHealth.................................... 11
3.3 WorldBankAidData .............................. 12
4 Epidemics and Human Capital Development 13
4.1 Meningitis Shocks and Human Capital Development . . . . . . . . . . . . . . 13
4.1.1 Mechanisms: The Role of WHO Epidemic Year Announcements . . . 15
4.2 OLSEstimates .................................. 16
4.3 Selective Migration and Inference Robustness . . . . . . . . . . . . . . . . . 19
4.4 Channels...................................... 21
5 Role of Health Aid: Evidence from World Bank Projects 24
5.1 How World Bank Projects are Approved and Funded . . . . . . . . . . . . . 24
5.2 WorldBankAidResults ............................. 25
6 Conclusion 26
A Appendix (For Online Publication) 48
A.1 Summary Statistics and Robustness . . . . . . . . . . . . . . . . . . . . . . . 52
A.1.1 WealthIndex ............................... 52
48
A.2 SelectiveMigration ................................ 54
A.2.1 Migration Estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
A.2.2 Empirical Test: Role of Selective Migration . . . . . . . . . . . . . . . 55
A.3 Meningitis Epidemics, Economic Activity and World Bank Aid . . . . . . . . 56
A.3.1 Results for Night Light Density . . . . . . . . . . . . . . . . . . . . . 58
A.3.2 Health Aid in Epidemic and Non-Epidemic Years . . . . . . . . . . . 60
A.3.3 Night Light Density as a Control . . . . . . . . . . . . . . . . . . . . 60
A.4 Health Expenditures and Aid in the Meningitis Belt . . . . . . . . . . . . . . 64
A.4.1 Domestic Government Effort . . . . . . . . . . . . . . . . . . . . . . . 65
List of Figures
1 Countries in the African Meningitis Belt (a), with districts in study region (b)
and locations of World Bank aid projects for countries and districts in study
regionoverstudyyears(c)............................ 28
2 Stochastic dominance: Child weight for age and height for age z-scores are
lower in meningitis shock districts on average . . . . . . . . . . . . . . . . . 29
3 Average child current health outcomes in meningitis shock and non-shock
districts. Share stunted and underweight and total vaccination is higher in
most years in meningitis shock districts. . . . . . . . . . . . . . . . . . . . . 29
A1 Mean weekly meningitis cases per district over study region, with epidemic
yearsspeciedinbrackets ............................ 52
A2 Selective Migration Tests: (a) OLS effect of meningitis shock on child health
in epidemic vs non-epidemic years with trimming of highest weight-for-age
(WFA z) in meningitis affected districts; (b) OLS effect of meningitis shock
on child health in epidemic vs non-epidemic years with trimming of highest
height-for-age (HFA z) in meningitis affected districts . . . . . . . . . . . . . 57
49
A3 World Bank project approval example snapshot . . . . . . . . . . . . . . . . 65
A4 Top 5 World Bank health and non-health projects funded by project title in
epidemic and non-epidemic years . . . . . . . . . . . . . . . . . . . . . . . . 66
List of Tables
1 SummaryStatistics................................ 30
2 Effect of meningitis shock on child current weight and height outcomes, at/near
birth (bcg, polio, dpt) versus non-at/near birth recommended (measles) child
vaccinations, total vaccinations and infant mortality . . . . . . . . . . . . . . 31
3 Effect of meningitis shock on child current weight and height outcomes, at/near
birth (bcg, polio, dpt) versus non-at/near birth recommended (measles) child
vaccinations, total vaccinations and infant mortality in epidemic versus non-
epidemicyears................................... 32
4 Robustness to marginal changes in shock cutoff: Effect of meningitis shock
(>0.5) on child current weight and height outcomes, at/near birth (bcg, polio,
dpt) versus non-at/near birth recommended (measles) child vaccinations, total
vaccinations and infant mortality in epidemic versus non-epidemic years . . . 33
5 Robustness to continuous measure: Effect of meningitis shock (continuous z-
score) on child current weight and height outcomes, at/near birth (bcg, polio,
dpt) versus non-at/near birth recommended (measles) child vaccinations, total
vaccinations and infant mortality in epidemic versus non-epidemic years . . . 34
6 Robustness: Effect of meningitis shock on child current weight and height
outcomes, at/near birth (bcg, polio, dpt) versus non-at/near birth recom-
mended (measles) child vaccinations, total vaccinations and infant mortality
in epidemic versus non-epidemic years . . . . . . . . . . . . . . . . . . . . . . 35
50
7 Placebo tests for Epidemic year: Effect of meningitis shock on child health
(weight-for-age WFA zand height-for-age HFA z)............... 36
8 Placebo tests for Epidemic year: Changing epidemic year cutoff . . . . . . . 37
9 Effect of meningitis shock on child current weight and height outcomes, at/near
birth (bcg, polio, dpt) versus non-at/near birth recommended (measles) child
vaccinations, total vaccinations and infant mortality by wealth status . . . . 38
10 Effect of meningitis shock on amount committed and disbursed to World Bank
aid projects by epidemic year and health project status . . . . . . . . . . . . 39
A1 WHO recommended vaccination schedule . . . . . . . . . . . . . . . . . . . . 52
A2 Balance on geographic and institutional characteristics . . . . . . . . . . . . 53
A3 Internal Migration Statistics for Selected Countries in the Meningitis Belt,
1988-1992, Source: Bocquier and Traore (1998) . . . . . . . . . . . . . . . . 55
A4 Effect of meningitis shock on economic activity in epidemic vs non-epidemic
years (Models: country x year FE and district specific time trends) . . . . . 59
A5 Effect of meningitis shock on night light density outcomes by World Bank aid
share of health projects, and total committed and disbursed aid . . . . . . . 61
A6 Effect of meningitis shock on child current weight and height outcomes, at/near
birth (bcg, polio, dpt) versus non-at/near birth recommended (measles) child
vaccinations, total vaccinations and infant mortality . . . . . . . . . . . . . . 62
A7 Effect of meningitis shock on child current weight and height outcomes, at/near
birth (bcg, polio, dpt) versus non-at/near birth recommended (measles) child
vaccinations, total vaccinations and infant mortality in epidemic versus non-
epidemicyears................................... 63
A8 Reduced Form Relationship Between Epidemic Year and Health Expenditures
for Meningitis Belt Countries, 1995-2008 . . . . . . . . . . . . . . . . . . . . 66
51
A9 Effect of meningitis epidemics on ODA aid flows committed to belt countries,
1995-2008 ..................................... 67
A10 Domestic government health expenditure . . . . . . . . . . . . . . . . . . . . 67
A.1 Summary Statistics and Robustness
Figure A1: Mean weekly meningitis cases per district over study region, with epidemic years
specified in brackets
Table A1: WHO recommended vaccination schedule
Vaccine Diseases Age
1 BCG tuberculosis at birth
2 Polio (OPV) polio at birth, 6, 10, 14 weeks
3 DPT diphtheria, pertussis, tetanus 6, 10, 14 weeks
4 Measles measles 9 months
A.1.1 Wealth Index
The wealth index is constructed from ownership of 10 assets using principal component anal-
ysis of asset ownership from the DHS. The assets include: bicycles, motorcycles, cars/trucks,
flush toilets, ventilated improved pit latrines, traditional pit latrines, electricity, radio, tv,
fridge. The wealth index is a 1 to 5 categorical variable where 1 is the poorest quintile and 5
52
Table A2: Balance on geographic and institutional characteristics
Panel A: Geographic Characteristics
Malaria Land Suitability Elevation Access to Rivers Distance to Sea Coast Distance to Capital Precipitation
(1) (2) (3) (4) (5) (6) (7)
Meningitis shock average 1.680 0.007 18.696 0.077 22.516 19.465 0.274
(3.214) (0.081) (51.375) (0.339) (57.331) (131.928) (0.279)
Mean of outcome 22.204 0.325 374.821 0.467 128.404 404.695 10.583
Observations 242 239 242 242 242 242 238
R20.576 0.503 0.554 0.094 0.322 0.250 0.495
Country FE Yes Yes Yes Yes Yes Yes Yes
Panel B: Geographic and Institutional Characteristics
Share Muslim Pastoral Centralization Index Centralization Dummy Diamond Petrol Temperature
(1) (2) (3) (4) (5) (6) (7)
Meningitis shock average 0.218 0.025 1.182 0.419 0.009 0.002 0.452
(0.149) (0.052) (0.867) (0.437) (0.100) (0.007) (0.470)
Mean of outcome 0.688 0.026 1.288 0.721 0.012 0.004 299.988
Observations 236 764 768 768 242 242 238
R20.536 0.191 0.078 0.055 0.092 0.025 0.545
Country FE Yes Yes Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors clustered by district in parentheses. Observations at the district level in all specifications except panel B for the centralization and pastoral outcomes,
where observations are districts intersected with Murdock ethnicity regions. ‘Meningitis shock average’ is the likelihood that a district is a meningitis shock district over the period of study. Land Suitability is land
suitability for agriculture from FAO data. Elevation is mean elevation in km from the Global Climate database. Distance to capital and seacoast in km. Malaria stability is from the malaria ecology index from
Kiszewski et al. (2004). Precipitation is in logs and Temperature is in Kelvin from the NASA MERRA-2 dataset. Share Muslim is based on DHS data. Access to Rivers is an indicator for whether a district has a
river running through it. Centralization Index is the level of precolonial centralization from Murdock ethnicity data (Murdock, 1967) and Centralization Dummy is an indicator that equals 1 if the index is greater
than 0 (following Archibong (2019)). Pastoralism dummy equals 1 if pastoralism was the primary contributor to livelihood in precolonial ethnic region from Murdock data. Petrol and Diamond are indicators equal
to 1 if the district has recorded deposits of petroleum and diamonds respectively from the PRIO dataset. ∗∗∗Significant at the 1 percent level, ∗∗Significant at the 5 percent level, Significant at the 10 percent level.
is the richest quintile. Lower wealth is an indicator that equals one if the household wealth
index is less than 3.
53
A.2 Selective Migration
To what extent does migration rationalize our results? We investigate the possibility that
unhealthy individuals (i.e., with low WFA z, low HFA z, etc) might have moved from areas
affected by meningitis to unaffected areas and as a result, unaffected areas experience low
economic outcomes relative to the affected areas. The dual, though prima facie less plausible,
statement is that more “healthy” individuals might have moved from areas unaffected by
meningitis to the affected areas and as result, unaffected areas experience low economic
outcomes. Thus, instead of assuming limited (selective) internal migration between districts
for identification, we relax this assumption and examine it as an alternative explanation for
our results.
A.2.1 Migration Estimates
We evaluate the extent of migration across districts to gauge its likely effects. Because
detailed micro data on internal migration over the entire sample period (1986 to 2008) is
absent, we provide estimates based on the ACMI (aggregate crude migration index) and net
migration rate (NMR) values calculated from 1988 to 1992 in Bocquier and Traor´e (1998).
In the demography literature, ACMI is a widely-used measure of internal migration and
captures the share of the population that has changed address averaged over a specified time
period. Specifically, the ACMI is a global average based on the specification:
CMIn=X
i
X
j6=i
Mij /X
i
Pi
where Mij is the total number of migrants (or migrations) between origin area i=1,...n
and destination area j=1,...n; and Piis the population of each area iat risk of migrating
(Bell et al., 2015; Bernard and Bell, 2018). The population assessed here is the population
over the age of 15 (Bocquier and Traor´e, 1998). The NMR measures the difference between
54
incoming and outgoing migrants in a particular locality.
Table A3 shows the ACMI and NMR (%) values, and indicates extremely low values.
Overall, ACMI averages at 0.09 while NMR averages at -0.72%. This means that just 9% of
the population report changing their place of residence within their country over the four-
year interval (1988 to 1992) with a net movement of -0.72%. The evidence suggests limited
internal migration in the study region.
Table A3: Internal Migration Statistics for Selected Countries in the Meningitis Belt, 1988-
1992, Source: Bocquier and Traore (1998)
Country ACMI (4-yr avg) NIMR ( %)
Capital city Principal towns Secondary towns Rural
Burkina Faso 0.03 1.86 0.29 -0.79 -0.09
Cote d’Ivoire 0.16 0.43 -2.24 -2.74 0.99
Guinea 0.05 1.21 -1.94 -2.14 -0.04
Mali 0.09 0.85 0.31 0.23 -0.19
Mauritania 0.08
Senegal 0.12 0.5 0.36 -0.6 -0.25
Niger 0.06 -0.06 0.91 -0.22 -0.04
West Africa (8) 0.09 0.8 -0.39 -1.04 0.06
Sample years 1988-1992 1988-1992 1988-1992 1988-1992 1988-1992
Notes: ACMI is the aggregate crude migration intensity ratio described in the text. NIMR is the net internal migration rate in percentages.
It is calculated for each region. Regional classification of ‘principal’ or ‘secondary’ towns differs for each country and is based on popu-
lation size. For Niger, principal towns are regional capital cities, and secondary towns are all remaining settlements of over 5000 people
(Beauchemin and Bocquier, 2004).
A.2.2 Empirical Test: Role of Selective Migration
To test our conjecture that (selective) migration is not driving the results, we conduct a series
of trimming exercises. We begin with the supposition that migration is indeed selective, and
then ask “what level of such selective migration would be needed to make our results insignif-
icant?”. We reclassify the districts as either meningitis affected (if the observed meningitis
cases are above the sample average) or unaffected (if the observed meningitis cases are below
the sample average) year to year. We then trim the outcomes using different migration rates
in increments of 5%. That is, we recursively drop the top 5%, 10%, 15%, ... of the data
55
with the highest outcomes- reflecting the most healthy individuals- only in the meningitis
affected districts. In each step, we re-estimate our baseline model, and continue the process
until the effects for our main interaction term,“Meningitis shock x Epidemic year”, become
insignificant.
Figure A2 shows the results. We focus on two main outcomes, WFA z and HFA
z25. WFA z and HFA z correlate strongly with the other child health outcomes (a simple
regression of WFA z and HFA z on the other health outcomes shows large and significant
correlations, p < 0.01). As shown, for WFA z and HFA z, a selective migration rate of 55%
is required to render our effects insignificant. . The coefficient signs remain unchanged
across all specifications. Our trimming exercise results suggest that migration would have
to, differentially, rise by at least 55% to explain the results, which is very unlikely given the
empirical evidence in Section A.2.1. This evidence is consistent with the fact that most of
the districts are rural where (selective) migration may be difficult to achieve. The results
are consistent with other papers showing a lack of selective migration in developing country
settings (Bazzi et al., 2016).
A.3 Meningitis Epidemics, Economic Activity and World Bank Aid
Meningitis shocks may affect economic activity directly through either income effects on
households, as discussed previously, or through their effects on triggering an inflow of aid
in declared epidemic years. We examine the relationship between meningitis shocks and
economic activity. Following the literature using night light density as a proxy for economic
activity (Henderson, Storeygard, and Weil, 2011; Michalopoulos and Papaioannou, 2013), we
use data on night light density from the National Oceanic and Atmospheric Administration
(NOAA) Defense Meteorological Satellite Program’s Operational Linescan System (DMSP-
OLS) to measure economic activity in the absence of detailed microlevel income estimates
25The results are consistent and available for other outcomes upon request.
56
a. b.
Meningitis shock
Epidemic year
Meningitis shock x Epidemic year
1.5 0 .5 1
OLS Estimates
0% Trimming
5% Trimming
10% Trimming
15% Trimming
20%
25%
30%
35%
40%
45%
50%
55%
60%
65%
70%
75%
80%
85%
90%
95% Trimming
Meningitis shock
Epidemic year
Meningitis shock x Epidemic year
1.5 0 .5
OLS Estimates
0% Trimming
5% Trimming
10% Trimming
15% Trimming
20%
25%
30%
35%
40%
45%
50%
55%
60%
65%
70%
75%
80%
85%
90%
95% Trimming
Notes: Figure plots the distribution of estimates under various trimming values.
Regressions (a total of 20) are estimated by OLS. Dependent variables are child health,
weight-for-age WFA z and height-for age HFA z, in (a) and (b) respectively. Meningitis
shock is z-score indicator based on district level mean. Models include full set of fixed
effects and district level linear time trends. Robust standard errors in parentheses clustered
by district. 90% confidence intervals are shown by horizontal lines, separately for each
regression. Color codes: blue denotes the baseline model (with no trimming), and red
denotes insignificant estimates for the main interaction term- showing the trimming level
that interaction term turns insignificant.
Figure A2: Selective Migration Tests: (a) OLS effect of meningitis shock on child health
in epidemic vs non-epidemic years with trimming of highest weight-for-age (WFA z) in
meningitis affected districts; (b) OLS effect of meningitis shock on child health in epidemic
vs non-epidemic years with trimming of highest height-for-age (HFA z) in meningitis affected
districts
57
for the study countries. Night light density data from the NOAA is available from 1992, and
we use data from 1992 to 2008 to match meningitis case data from our study region. Since a
notable fraction of the district level observations take on the value of zero, following previous
literature, we use the log of night light density, adding a small number (ln (0.01+ Light
Density)) as our measure of night light density (Michalopoulos and Papaioannou, 2013).
The log transformation allows us to use all observations and account for outliers in the
luminosity data (Michalopoulos and Papaioannou, 2013). In alternate specifications, we use
different transformations of the night light density measure, like the arcsine transformation.
While the results using the arcsine transformation are qualitatively similar, they are not
robust, and hence, results using the log night light density measure should be interpreted
with caution.
A.3.1 Results for Night Light Density
Table A4 reports estimates from Equation 2 with the night light density outcome. First,
we interpret the results from the country-year FE model. On average, meningitis shocks
reduce economic activity, as measured by night light density, by 6.5% as shown in column
(1). The effect is nonlinear, as shown in the fully specified models in columns (2). Meningitis
shocks increase economic activity by around 17.1% in epidemic years and reduce economic
activity by 14.2% in non-epidemic years. The effect of meningitis shocks during epidemic
years is effectively reversed, with an increase in economic activity of up to 2.9% in meningitis
shock districts during declared epidemic years. The results are nearly identical in the linear
time trend specification from Equation 2, and the estimates are largely stable, if slightly
underpowered, as shown in columns (3) and (4).
We can benchmark these nightlight density results to GDP growth rate figures using
a simple back of the envelope calculation based on estimates from Henderson, Storeygard,
and Weil (2011) where a 1% increase in nightlight density increases GDP growth rates by
58
about 0.3% in low and middle income countries. Back of the envelope calculations show that
meningitis shocks can reduce GDP growth rates by between 2% and 4.3% in the absence of
a WHO epidemic declaration.
The results are striking, in that although the average effect of meningitis shocks is
negative, there is significant heterogeneity in the effects of these shocks depending on whether
or not the WHO declares an epidemic year. Given the high share of health expenditure
sourced from donor aid in the majority of the study countries as discussed in Section 3, a
major mechanism explaining this result may be an influx of disaster aid when the WHO
declares an epidemic year.
Table A4: Effect of meningitis shock on economic activity in epidemic vs non-epidemic years
(Models: country x year FE and district specific time trends)
Log Night Light Density
(1) (2) (3) (4)
Meningitis shock 0.075∗∗ 0.0650.142∗∗ 0.142
(0.033) (0.036) (0.064) (0.088)
Meningitis shock
x Epidemic year 0.171∗∗ 0.159.
(0.082) (0.099)
Mean of outcome -2.741 -2.741 -2.741 -2.741
District FE Yes Yes Yes Yes
Country x year FE No Yes Yes NA
Year FE NA NA NA Yes
Linear time trends (D) NA NA NA Yes
Observations 1,141 1,141 1,141 1,141
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district.Two models estimated
with country-year FE in columns (1) to (3) and district specific time trends in column (4) as described in text. The Epidemic
year coefficient is omitted in the model with country x year FE. The Epidemic year coefficient in column (4) with district
time trends is -0.030 and insignificant at conventional levels. Dependent variables are log night light density described in
text from 8 African countries from 1992 to 2008. Meningitis shock is Z score indicator based on district level mean as
described in text. ∗∗∗ Significant at the 1 percent level, ∗∗ Significant at the 5 percent level, Significant at the 10 percent
level. .Significant near 10 percent level with p0.1.
59
A.3.2 Health Aid in Epidemic and Non-Epidemic Years
Although we do not have detailed data on total amounts disbursed to health projects in each
year, we can test whether or not meningitis shock districts that receive a greater share of
health aid projects experience increases in economic activity directly by re-estimating the
models in Equation 2 with country-year FE to account for the small sample, and interacting
the meningitis shock variable with the share of health projects funded in each district. The
results are shown in Table A526. Meningitis shock districts that receive a greater share of
health aid projects and more health aid committed and disbursed experience an increase in
their economic activity as measured by night light density and shown in columns (1)-(3).
The effect is driven by health specific aid not non-health aid as shown in columns (4) and
(5).
A.3.3 Night Light Density as a Control
While the analysis in Section A.3.1 and Section A.3.2 assumes night light density is an
outcome of meningitis shock, we can test a different hypothesis, where we control for night
light density as a further robustness check. The results in Table A6 and Table A7 remain
largely unchanged, with similar estimates as in the main results in Section 4. The results here
should be interpreted with caution, since as mentioned in the previous section, a plausible
hypothesis is that night light density/economic activity is an outcome of meningitis shocks.
26There is not enough power for a triple interaction or split sample approach including the declared
epidemic year.
60
Table A5: Effect of meningitis shock on night light density outcomes by World Bank aid
share of health projects, and total committed and disbursed aid
Log Night Light Density
(1) (2) (3) (4) (5)
Meningitis shock 0.0940.1030.1030.767 0.153
(0.058) (0.061) (0.061) (1.578) (0.310)
Share health 0.055
(0.222)
Comm. health 0.130
(0.117)
Disb. health 0.131
(0.117)
Comm. non-health 0.002
(0.006)
Disb. non-health 0.002
(0.007)
Meningitis shock x Share health 0.188
(0.095)
Meningitis shock x Comm. health 0.009
(0.005)
Meningitis shock x Disb. health 0.009
(0.005)
Meningitis shock x Comm. non-health 0.010
(0.006)
Meningitis shock x Disb. non-health 0.007
(0.006)
Mean of outcome 3.056 3.056 3.056 3.056 3.056
District FE Yes Yes Yes Yes Yes
Country x year FE Yes Yes Yes Yes Yes
Observations 147 147 147 147 147
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district. Dependent variable is Log night light density described
in text from 8 African countries. Meningitis sho ck is Z score indicator based on district level mean as described in text. Results qualitatively similar with
district specific time trends (D). ∗∗∗Significant at the 1 percent level, ∗∗Significant at the 5 percent level, Significant at the 10 percent level.
61
A.4 Health Expenditures and Aid in the Meningitis Belt
Table A8 reports results on the effects of meningitis epidemics on private health expenditures.
There is a significant increase in prepaid private health expenditures27 in the 20 meningitis
belt study countries between 1995 and 2008. Domestic government health spending, in
contrast, remains unchanged in response to epidemics. This is perhaps unsurprising given
that government health spending accounted for just over 23% of health spending among
meningitis belt countries, while out of pocket expenditures made up 47% of total health
spending as of 2017 by World Bank estimates. Meningitis epidemics are a notable negative
income shock to households in the belt. Given that these shocks pose a significant private
cost to households and the fact that 23% of health spending in the belt comes from external,
donor sources, do these donors/lenders respond with increased financing to belt countries
during epidemics?
Table A9 reports results on the effects of epidemics on ODA aid flows committed to
meningitis belt countries. There is no effect of epidemic year declarations on total aid com-
mitted to belt countries during the epidemic year as shown in column (2). The share of
aid committed to health or to infectious disease control in particular is not significantly
associated with epidemic year declarations as shown in column (1). On the other hand, epi-
demic year declarations are strongly positively associated with the total amount committed
to infectious disease control and the share of infectious disease spending in total aid com-
mitted in the following year as shown in column (3) and column (4) of Table A9. National
government donor aid agencies are very slow to respond to epidemics in recipient countries.
Additionally, there is no increase in overall aid committed in the following year, suggesting
targeted increases in infectious disease spending only and potential crowd-out of non-health
spending following an epidemic year.
27Prepaid private spending includes private insurance and non-governmental agency spending.
64
In contrast, international financial organizations like the World Bank are quicker to
respond to epidemic declarations with crisis financing as shown in Table 10. The World
Bank funds more health projects during epidemic years, and increases the total amount
committed and disbursed to countries during the epidemic year. The results do not show
the same lag in funding from the Bank as in the national government donor agencies. There is
similar crowd-out, with World Bank aid funding distributing away from non-health projects
towards health projects.
A.4.1 Domestic Government Effort
We explore the possibility that the reversal in meningitis shock effects following WHO epi-
demic declarations may be driven by national governments domestic efforts. We define
government effort as potential investments in the health sector. We draw on country-level
panel data on health from World Bank’s World Development Indicators to derive two alter-
native measures of government’s domestic health sector effort, which include (i) domestic
general government health expenditure (% of current health expenditure) and (ii) domestic
general government health expenditure (% of general government expenditure). Using these
as outcomes, we estimate a modified version of the baseline regression model to test for the
potential role of government effort in mitigating the epidemic effect. Table A10 shows the
results and indicates no meaningful evidence of governments domestic efforts/investment.
Figure A3: World Bank project approval example snapshot
65
Figure A4: Top 5 World Bank health and non-health projects funded by project title in
epidemic and non-epidemic years
Table A8: Reduced Form Relationship Between Epidemic Year and Health Expenditures for
Meningitis Belt Countries, 1995-2008
Panel: Prepaid Private Spending (PPP) and Government Health Spending (GHES)
PPP/THE PPP/GDP PPP/CAP GHES/THE GHES/GDP GHES/CAP
(1) (2) (3) (4) (5) (6)
Epidemic Year 0.0050.0003∗∗ 0.455∗∗ 0.014 0.001 1.471
(0.003) (0.0001) (0.198) (0.016) (0.001) (1.136)
Mean of outcome 0.038 0.002 2.510 0.285 0.015 22.626
Observations 107 107 107 107 107 107
R20.970 0.938 0.975 0.810 0.827 0.893
Country FE Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses. Observations are 20 meningitis belt countries for which data is available
over 1995 to 2008 including: Benin, Burkina Faso, Cameroon, CAR, Cote d’Ivoire, DRC, Eritrea, Ethiopia, Gambia, Ghana, Guinea, Guinea Bissau,
Kenya, Mali, Mauritania, Niger, Nigeria, Senegal, Sudan, and Togo. CAP is per capita. GDP is per GDP in 2015 USD PPP. Country and year fixed
effects included in all specifications. Source: Global Burden of Disease Health Financing Collaborator Network. ∗∗∗ Significant at the 1 percent level,
∗∗Significant at the 5 percent level, Significant at the 10 percent level.
66
Table A9: Effect of meningitis epidemics on ODA aid flows committed to belt countries,
1995-2008
Concurrent Spending, t Spending, t+1
Infectious/Total Comm. Total Comm. Infectious Infectious/Total Health/Total Comm. Total
(1) (2) (3) (4) (5) (6)
Epidemic Year 0.003 0.033 0.895∗∗∗ 0.005∗∗ 0.008 0.153
(0.003) (0.091) (0.331) (0.003) (0.010) (0.100)
Mean of outcome 0.009 20.430 14.446 0.009 0.064 20.420
Observations 78 112 60 60 91 91
R20.609 0.920 0.818 0.557 0.406 0.950
Country FE Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses. Observations are 20 meningitis belt countries for which data is available
over 1995 to 2008 including: Benin, Burkina Faso, Cameroon, CAR, Cote d’Ivoire, DRC, Eritrea, Ethiopia, Gambia, Ghana, Guinea, Guinea Bissau,
Kenya, Mali, Mauritania, Niger, Nigeria, Senegal, Sudan, and Togo. C oncurrentS pending is same year spending in columns (1) and (2). S pending,t+1
is spending in the following year. Comm. Total is log (total committed real (2010) dollars). Comm. Infectious is log (1+ total committed real dollars
to infectious disease control). Source: OECD CRS data ∗∗∗Significant at the 1 percent level, ∗∗Significant at the 5 percent level, Significant at the 10
percent level.
Table A10: Domestic government health expenditure
Govt. health exp. (% of current health exp.) Govt. health exp. (% of general govt. exp.)
(1) (2) (3) (4) (5) (6)
Meningitis shock 0.448 0.129 0.089 0.049
(0.303) (0.601) (0.057) (0.116)
Epidemic year 0.634∗∗ 0.7250.130∗∗ 0.156∗∗
(0.268) (0.418) (0.051) (0.079)
Meningitis shock
x Epidemic year 0.018 0.014
(0.424) (0.082)
Constant 29.900∗∗∗ 29.550∗∗∗ 29.960∗∗∗ 6.496∗∗∗ 6.423∗∗∗ 6.511∗∗∗
(0.296) (0.256) (0.359) (0.110) (0.106) (0.118)
Observations 15,032 15,032 15,032 15,032 15,141 15,141
R20.915 0.915 0.915 0.916 0.915 0.916
Country FE Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses. Regressions estimated by OLS. Dependent variables are two alternative
measures of government’s domestic health sector effort: (i) domestic general government health expenditure (% of current health expenditure) and (ii)
domestic general government health expenditure (% of general government expenditure). Country-level panel data on (i) and (ii) from World Bank’s World
Development Indicators merged with WHO national epidemic declarations and district-level measures on Meningitis shock for the 8 African countries.
Year FE are survey year fixed effects. Meningitis shock is z-score indicator based on district-level mean and equal to 1 if the z-score is greater than 0.
∗∗∗Significant at the 1 percent level, ∗∗Significant at the 5 percent level, Significant at the 10 percent level.
67
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
Introduction The global abrupt progression of the COVID-19 pandemic may disrupt critical life-saving services such as routine immunization (RI), thus increasing the susceptibility of countries to outbreaks of vaccine-preventable diseases (VPDs). Being endemic to several infectious diseases, Lebanon might be at increased risk of outbreaks as the utilization of RI services might have deteriorated due to the pandemic and the country’s political unrest following the October 2019 uprising. The aim of this study was to assess the changes in the utilization of RI services in both the public and private sectors following the COVID-19 pandemic. Methods A self-administered cross-sectional survey was completed electronically, in April 2020, by 345 private pediatricians who are registered in professional associations of physicians in Lebanon and provide immunization services at their clinics. Means of the reported percentages of decrease in the utilization of vaccination services by pediatricians were calculated. As for the public sector, an examination of the monthly differences in the number of administered vaccine doses in addition to their respective percentages of change was performed. Adjustment for the distribution of RI services between the sectors was performed to calculate the national decrease rate. Results The utilization of vaccination services at the national level decreased by 31%. In the private sector, immunization services provision diminished by 46.9% mainly between February and April 2020. The highest decrease rates were observed for oral poliovirus vaccine (OPV) and hepatitis A, followed by measles and pneumococcal conjugate vaccines. The number of vaccine doses administered in the public sector decreased by 20%. The most prominent reductions were detected for the OPV and measles vaccines, and during October 2019 and March 2020. Conclusion The substantial decrease in the utilization of RI as a result of the COVID-19 pandemic requires public health interventions to prevent future outbreaks of VPDs.
Article
Full-text available
This paper provides a theoretical framework to understand the tendency of non governmental organisations (NGOs) to cluster and the circumstances under which such clustering is socially undesirable. NGOs compete through fundraising for donations and choose issues to focus their projects on. Donors have latent willingness-to-give that may differ across issues, but they need to be ”awakened” to give. Raising funds focusing on the same issue creates positive informational spillovers across NGOs. Each NGO chooses whether to compete in the same market (clustering) with spillovers, or to face weaker competition under issue specialization. We show that equilibrium clustering is more likely to occur when the share of multiple-issue donors is relatively large, and when the fundraising technology is sufficiently efficient. Moreover, this situation is socially inefficient when the cost of fundraising takes intermediate values and the motivation for donors’ giving is relatively high. We illustrate the mechanisms of the model with several case studies.
Article
Full-text available
Backgrounds: Up to February 16, 2020, 355 cases have been confirmed as having COVID-19 infection on the Diamond Princess cruise ship. It is of crucial importance to estimate the reproductive number (R0) of the novel virus in the early stage of outbreak and make a prediction of daily new cases on the ship. Method: We fitted the reported serial interval (Mean and standard deviation) with a gamma distribution and applied "earlyR" package in R to estimate the R0 in the early stage of COVID-19 outbreak. We applied "projections" package in R to simulate the plausible cumulative epidemic trajectories and future daily incidence by fitting the data of existing daily incidence, a serial interval distribution, and the estimated R0 into a model based on the assumption that daily incidence obeys approximately Poisson distribution determined by daily infectiousness. Results: The Maximum-Likelihood (ML) value of R0 was 2.28 for COVID-19 outbreak at early stage on the ship. The median with 95% confidence interval (CI) of R0 values was 2.28 (2.06-2.52) estimated by the bootstrap resampling method. The probable number of new cases for the next ten days would gradually increase, and the estimated cumulative cases would reach 1514 (1384-1656) at the tenth day in the future. However, if R0 value was reduced by 25% and 50%, the estimated total number of cumulative cases would be reduced to 1081 (981-1177) and 758 (697-817), respectively. Conclusion: The median with 95% CI of R0 of COVID-19 was about 2.28 (2.06-2.52) during the early stage experienced on the Diamond Princess cruise ship. The future daily incidence and probable outbreak size is largely dependent on the change of R0. Unless strict infection management and control are taken, our findings indicate the potential of COVID-19 to cause greater outbreak on the ship.
Article
Many studies use shift-share (or “Bartik”) instruments, which average a set of shocks with exposure share weights. We provide a new econometric framework for shift-share instrumental variable (SSIV) regressions in which identification follows from the quasi-random assignment of shocks, while exposure shares are allowed to be endogenous. The framework is motivated by an equivalence result: the orthogonality between a shift-share instrument and an unobserved residual can be represented as the orthogonality between the underlying shocks and a shock-level unobservable. SSIV regression coefficients can similarly be obtained from an equivalent shock-level regression, motivating shock-level conditions for their consistency. We discuss and illustrate several practical insights of this framework in the setting of Autor et al. (2013), estimating the effect of Chinese import competition on manufacturing employment across U.S. commuting zones.
Article
Zimbabwe’s catastrophic cholera outbreak of 2008/09 resulted in an unprecedented 100,000 cases and nearly 5,000 deaths. In the aftermath of the epidemic, questions of suffering and death and of rescue, relief, and rehabilitation have persisted in on-going processes of meaning-making through which people come to terms with the epidemic as a ‘man-made’ disaster. Based on extensive fieldwork, I examine the views of residents in Harare’s high-density townships that were epicentres of the disease. I argue that cholera was experienced by township residents as many crises at the same time. It was not only a public health crisis but also a political–economic crisis, a social crisis as well as a crisis of expectations, history and social identity. As such, I argue that the cholera outbreak was intensely generative of political subjectivities that reveal important shifts in the fraught relations between state and society in Zimbabwe’s urban politics. Finally, I argue that the government’s perceived causal role in, and failure to respond to, the cholera outbreak occasioned intense public outrage among township residents, which speaks to a much deeper aspiration for substantive citizenship based on political rights, social recognition, and access to high-quality public services delivered by a robust, responsible state.
Article
The Bartik instrument is formed by interacting local industry shares and national industry growth rates. We show that the typical use of a Bartik instrument assumes a pooled exposure research design, where the shares measure differential exposure to common shocks, and identification is based on exogeneity of the shares. Next, we show how the Bartik instrument weights each of the exposure designs. Finally, we discuss how to assess the plausibility of the research design. We illustrate our results through two applications: estimating the elasticity of labor supply, and estimating the elasticity of substitution between immigrants and natives. (JEL C51, F14, J15, J22, L60, R23, R32)
Book
This book explores how population mobility varies among the countries of Asia. While much attention has been given to international migration, movement within countries is numerically much more significant. Coupling innovative methods developed in the global IMAGE project with the contextual knowledge of experts on 15 Asian countries, the book measures and explains how people across Asia differ in the probability of changing residence, the ages at which they move, and the impact of these migrations on the distribution of human settlement within each country. It demonstrates how stage of economic development, coupled with historical events, local contingencies, cultural norms, political frameworks, and the physical environment shape human migration. By using rigorous statistics in a robust comparative framework, this book provides a clear understanding of contemporary migration in Asia for students and academics, and a valuable resource for policy-makers and planners in Asia and beyond.
Book
Cambridge Core - African Studies - The Political Life of an Epidemic - by Simukai Chigudu